SELECTION IN BROILERS FOR TOLERANCE TO GROWTH . DEPRESSANTS IN RAW SOYBEANS Thesis for the Degree of Ph. D. MICHIGAN STATE UNIVERSITY PRAFULLA KUMAR PAN! ‘ 1972 .‘ r-c'r 51* J' L! B RAR Y Michigan State Univcrn'ty This is to certify that the thesis entitled Selection in Broil ers for Tolerance to Growth-Depressan ts in Raw Soybeans presented by Profulla Kumar Pani has been accepted towards fulfillment of the requirements for JILL degree uni-W5 cience //%L‘ C ‘ ((\"\~/€)\ Major professor Meg/.2 4 [7.2 0-169 I ABSTRACT SELECTION IN BROILERS FOR TOLERANCE TO GROWTH-DEPRESSANTS IN RAW SOYBEAN By Prafulla K. Pani Records of 1,997 White Plymouth Rock chicks produced in the years 1965 through 1968 from 42 males and 204 females were analyzed following four generations of bi-directional selection, for body weight at four weeks of age, of chicks fed a ration containing raw soybeans. Changes in an unselected trait, the body size at the same age for chicks of the same genetic group reared in an environment where they received cooked rather than raw soybeans in their ration, were analyzed to measure the effect of selection in one environment on the same trait in a different environment. In the line selected for high weight 784 chicks were produced by 22 sires and 92 dams; in the line selected for low weight 1,213 chicks were obtained from 20 sires and 112 dams. Data were analyzed within generations by least squares procedures. A mixed model was assumed with rearing environment, selected line and sex as fixed factors; sires within lines and dams within sires as random factors. The effect of positive selection was highly significant (P < .01) for all generations with a linear increase of 14.2 gms per generation. Prafulla K. Pani Similarly, a declining trend between the difference for the effect of rearing environment was observed from the second generation to the fourth generation {-14.7 gms per generation). However, considering all four generations, the corresponding estimate was —4.0 gms. Inter- action of line (genotype) and rearing environment was highly significant (P < .01) for all generations, ranging from 6.0 i 2.3 gms to 14.03: 2.4 gms. For chicks receiving the ration containing raw soybeans, mean changes in body weight per generation were —2.7 1:4.1 gms and -36.8;: 2.8 gms, respectively, in lines selected for high and low weight. Thus, by selection, the two lines were separated in body size by a significant difference (29.3 gms per generation) (P < .01). Possible causes of difference between responses to selection in the high and low lines were investigated. Correlated Response As a correlated trait, body weight at four weeks, for the chicks of the high and low line reared in a dietary environment in which they received a ration containing cooked soybeans, declined after the second generation. Reproductive fitness as a correlated response to selection to the high and low body size of the chicks did not decline over generations. Genetic Correlation The estimate of genotype-environment interaction obtained by Yamada's method, was used to compute the genetic correlation between a sire's ability for producing chicks to be reared in the two respective Prafulla K. Pani dietary environments. The estimates were 0.78 i 0.13 and 0.86 :_0.09 for the high and low line, respectively. Heritability Estimate Heritability estimates were calculated from full—sib intraclass correlations within rearing environment and line. The estimate had large sampling variation but no consistent difference was observed be- tween estimates obtained in high and low lines or in different environ- ments. Therefore, a pooled estimate was calculated within rearing environment (0.36 :_0.09). The average estimate of realized herita- bility agreed with the heritability estimated from full—sib intraclass correlations. Interaction of Sex and Rearing Environment Proportionally, males did not grow as well as the females when fed the ration containing raw soybeans. However, with a ration con- taining cooked soybeans, the males grew more rapidly than the females. Interaction of Sex and Line This interaction was not significant in any of the generations, i.e., line differences were consistent in both sexes. _—__————\*’_—’- dietary . for the : .. . .'.o; C 0.54. “‘7; . 8.; v Kr»'::_[' ..‘y.' Clint). ali: Ccrre.a:i. Q Q}- gugfr t \fi_ Q a“; 2 --f “‘ “ts. .7 Prafulla K. Pani dietary environments. The estimates were 0.78 :_0.13 and 0.86 :_0.09 for the high and low line, respectively. Heritability Estimate Heritability estimates were calculated from full—sib intraclass correlations within rearing environment and line. The estimate had large sampling variation but no consistent difference was observed be- tween estimates obtained in high and low lines or in different environ- ments. Therefore, a pooled estimate was calculated within rearing environment (0.36 :_0.09). The average estimate of realized herita- bility agreed with the heritability estimated from full-sib intraclass correlations. Interaction of Sex and Rearing Environment Proportionally, males did not grow as well as the females when fed the ration containing raw soybeans. However, with a ration con- taining cooked soybeans, the males grew more rapidly than the females. Interaction of Sex and Line This interaction was not significant in any of the generations, i.e., line differences were consistent in both sexes. SELECTION IN BROILERS FOR TOLERANCE TO GROWTH-DEPRESSANTS IN RAW SOYBEANS By Prafulla Kumar Pani A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Poultry Science 1972 u- 3. . t. ... I .t t. .... .. r. .6 .. . o s .0 . “a an .n u 5 st. 8 .s A .. a t& .9; t e F. 9‘ w.» L“ r. U C _._ . I .L ..T. C at 0 Lu U T. .q r. S .4 C .7. t. S C S I a. .1 e r T. x. .I t 3 a .L I t st 9 C t S 0 \w r d u .. . . E e r U C a a r CA a S .. c e a. 2 U r q 4 .n C an T. (1 U E r .l .2 C l .1 Wu L. o; L... e d. .3 f. 35 LL. . . n. . a: 3 ts e : a r 5 ~ a C .J 2* u f. O 7. my my Lt” AL flu Pg . as ‘1 CL 3 .L .l 2 . .1 Dr v . a t C. .I G 2 t D. C n II \IIII. [III (III III ACKNOWLEDGEMENTS The author is extremely indebted to Dr. T. H. Coleman, the author's academic advisor, for his guidance, encouragement and assis- tance during the present work. His patience and generous counseling in and out of the classroom will be an everlasting memory. A sincere appreciation is extended to Dr. W. T. Magee who took maximum of his time with the author while preparing this manuscript and without his patient understanding it would have been unsuccessful. To Dr. J. L. Gill who was all the time happy to solve the author's problems no matter how busy he was; the author is very much thankful. To the other member of the graduate committee, Dr. C. J. Flegal, the author's sincere appreciation is extended for reviewing and criticizing the nutritional portion of this manuscript. The author is especially indebted to the Department of Poultry Science of Michigan State University and Dr. H. C. Zindel, the chairman, for facilities and financial assistance provided in the form of a Graduate Assistantship. The author is also grateful to all other members of the staff, employees and graduate students in the depart- ment for their support, help and companionship. The author is also thankful to Dr. L. Barton for initiating the program under the guidance of Dr. T. H. Coleman. The author is extremely sorry to point out the sudden sad demise of his father who could not be able to live a little longer to share the ii happiness achieve 1' of this 5:? prays Geri sent and : happiness in this work. His unending inspiration for coming abroad to achieve this degree was the kindled light for the author in the progress of this graduate program. The author is extremely indebted to him and prays God for his soul to be peaceful. Above all, the author is very much thankful to his understanding wife Si-lu (Sailabala Pani), who, by her patience, continual encourage- ment and moral support made this work a grand success. iii F—‘ifi ti r—d D l U) r *l (J ll) 91 r.) T" k“ t .\‘., A.‘. 9— - VJ- ’ i L" f" " m 3 r0 2 :7: (It '1 (7 ) r1 1 p D r4 p—e H 'Y) n) : .l "9‘ ‘-‘ '7 ; i’t‘e '1 '1 m 0 U70 () "1 0" ”1 ‘°‘"-~ ’o. (‘9 "1 .I . . a» -.. as...‘. I I" fl y" D‘- r o A-.'..' ,4 (I: p .4 a» r) 1 TABLE OF CONTENTS Page LIST OF TABLES . . . . . . . . . . . . . . . . . . . . . . . . . vi LIST OF FIGURES . . . . . . . . . . . . . . . . . . . . . . . . viii INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . 1 OBJECTIVES . . . . . . . . . . . . . . . . . . . . . . . . . . 3 REVIEW OF LITERATURE . . . . . . . . . . 4 Growth Depressing Factors in Raw Soybean . . . . . . . . . . 4 Trypsin and Chymotrypsin Inhibitors . . . . . . . . . . . . . 4 Hemagglutinins . . . . . . . . . . . . . . . . . . . . . . . . 5 Saponins . . . . . . . . . . . . . . . . . . . . . . . . 6 Pancreatic Enlargement . . . . . . . . . . . . . . . . . . . . 6 Loss of Endogenous Nitrogen . . . . . . . . . . . . . . . . . 7 Metabolic Effects . . . . . . . . . . . . . . . . 7 Impaired Intestinal Digestion and Absorption . . . . . . 8 Influence of Other Amino Acids . . . . . . . . . . . . . . 8 Heritability . . . . . . . . . . . . . . . . . . . . . . 9 "Correlated Heritability" of Unselected Trait . . . . . . . . 11 Genotype x Environment Interaction . . . . . . . . . . . . . . 13 Genetic Correlation . . . . . . . . . . . . . . . . . . . . . 16 Correlated Response . . . . . . . . . . . . . . . . . . . . . 17 Sex Linkage and Sex Effects . . . . . . . . . . . . . . . . . 20 STATISTICAL PROCEDURE . . . . . . . . . . . . . . . . . . . . . 22 Genetic Correlation . . . . . . . . . . . . . . . . . . . . . 28 Standard Error of the Genetic Correlation . . . . . . . . . . 34 Estimation of "Heritability" . . . . . . . . . . . . . . . . . 34 Heritability Estimate from Intra—class Correlation Coefficient . . . . . . . . . . . . . . . . . . 36 Expected Phenotypic Correlation Between Selected and Correlated Traits . . . . . . . . . . . . . . . 37 MATERIALS AND METHODS . . . . . . . . . . . . . . . . . . . . . 41 iv _“',....V .Jb 9‘ ‘ h-v-"m‘ .\ s... ~Q - Ix...a .- . \ g.—qr\ .ar-O..Aus Va I N‘V Ah... uu-C-E" ‘ . (”I 7- F ..\.!K Y‘t.‘ ' F; etic q u {0" 55;. SEIQPP TABLE OF CONTENTS Page LIST OF TABLES . . . . . . . . . . . . . . . . . . . . . . . . . vi LIST OF FIGURES . . . . . . . . . . . . . . . . . . . . . . . . viii INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . 1 OBJECTIVES . . . . . . . . . . . . . . . . . . . . . . . . . . 3 REVIEW OF LITERATURE . . . . . . . . . . . . . 4 Growth Depressing Factors in Raw Soybean . . . . . . . . . 4 Trypsin and Chymotrypsin Inhibitors . . . . . . . . . . . . 4 Hemagglutinins . . . . . . . . . . . . . . . . . . . . . . . . 5 Saponins . . . . . . . . . . . . . . . . . . . . . . . . . . 6 Pancreatic Enlargement . . . . 6 Loss of Endogenous Nitrogen . . . . . . . . . . . . . . . . . 7 Metabolic Effects . . . . . . . . . . . . . . . . . 7 Impaired Intestinal Digestion and Absorption . . . . . . 8 Influence of Other Amino Acids . . . . . . . . . . . . . 8 Heritability . . . . . . . . . . . . . . . . . . 9 "Correlated Heritability" of Unselected Trait . . . . . . . . 11 Genotype x Environment Interaction . . . . . . . . . . . . . . 13 Genetic Correlation . . . . . . . . . . . . . . . . . . . . . l6 Correlated Response . . . . . . . . . . . . . . . . . . . . . 17 Sex Linkage and Sex Effects . . . . . . . . . . . . . . . . . 20 STATISTICAL PROCEDURE . . . . . . . . . . . . . . . . . . . . . 22 Genetic Correlation . . . . . . . . . . . . . . . . . . 28 Standard Error of the Genetic Correlation . . . . . . . . . . 34 Estimation of "Heritability" . . . . . . . . . . . . . . . . . 34 Heritability Estimate from Intra-class Correlation Coefficient . . . . . . . . . . . . . . . . . . 36 Expected Phenotypic Correlation Between Selected and Correlated Traits . . . . . . . . . . . . . . . 37 MATERIALS AND METHODS . . . . . . . . . . . . . . . . . . . . . 41 iv .h 7'. .a..‘ .- .anwon [I GENERAL EXPERIMENTAL PROCEDURE cow» . Base Population . Nutritional Rearing Environment . Selection and Mating . . Incubating and Brooding RESULTS AND DISCUSSION . Genotype x Environmental Interaction Asymmetry of Response Population Parameter Heritability, Realized O Heritability and Correlated Heritability . Correlated Response Correlated Response in Reproductive Fitness Genetic Correlation Sex Effect and Sex by Rearing Environment Interaction . Expected Phenotypic Correlation Between the 0 Selected and Correlated Traits . APPLICATION . . . Genetic Correlation CONCLUSIONS . . SUMMARY . . . . . LITERATURE CITED . Page 45 45 45 48 50 52 56 71 74 82 83 86 96 101 103 110 111 112 118 II I IIr I I I III II I n... “M .- v. 0» I... an sh. .L. e p . ..\ u. .. . s.‘ we. 10.. p. . e . v. v. ..... ... v“ a... . . . . I a 2. E .5. P. a s a . .1 1. G . .t r. E 5 new v... :1 Ox 3 .n. I. F v. a f Cw o a c s “a I». “L H. n. VD.“ «In A,“ I). 0 a 9 o .3 14 S 6 7 DC. . . . . a 9 0 7.4 2 TI. 1 1i l rq 10. ll. 12. LIST OF TABLES Expected Mean Square (for ANOVA Between Lines and Between Rearing Environments) . . . . . . . Expected Mean Square (ANOVA between Rearing Environment Within Line). . . . . . . . . . . . Expected Mean Square (Robertson, 1959). . . . . . Experimental Rations Used for the Chicks From One-Day-Old Until Four Weeks of Age . . . . . . Rate of Gain in Body Size per Day by Week for Different Line x Rearing Environments . . . Mean, Standard Deviation and Coefficient of Variation for Four-Week Body Weights (gms) by Lines and Generations . . . . . . . . . . . Regression Coefficients with S.E. for Various Variables Contributing Effects to the Mean for Four-Week Body Weight . . . . . . . . . . . Mean Body Weights (gms) Adjusted for Line, Sex, Rearing, Environment and Their Interaction Effects at 4 Weeks of Age by Generation for the High and Low Line Chicks Receiving Raw or Cooked Soybean in Their Ration . . . . . . . . ANOVA of Four-Week Body Weights (Between Lines and Between Environments) by Generation . . . . Expected and Observed Cumulative Response for Four-Week Body Weights by Lines and Generations O I O I O O O I O I O O O O O O O O Heritability Estimates for Four-Week Body Weight Analysis of Variance of Heritability Estimates for Four-Week Body Weights from Full-Sib Correlation . . . . . . . . . . . . . . . . . . vi Page 23 29 33 46 53 55 57 58 59 72 75 78 I" A“_\ 13. 15. o 0; FL YA Fl. 8 an .n A wbu (a . . 6 WI ‘IA 1 mu" Table 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. Selection Differential, Response and Realized Heritability for Four-Week Body Weight for Line by Generation . . . . . . . . . . . . . Mean Percentage of Fertility, Hatchability and Hatching Abnormalities (Adjusted for Line Effect) as Correlated Response to Selection of Body Size at Four Weeks of Age for the High and Low Lines Over Generations . . . . . ANOVA Between Rearing Environments Within Line for 4-Week Body Weight by Generations . . . . Estimated Components for Computation of Genetic Correlation by Various Methods . . . . . . . Sire-Progeny Unadjusted Means for Body Weight in Grams at Four Weeks of Age by Lines, Generations and Rearing Environment . . . . . . . . . . . Estimated Pooled Genetic Correlation over Generations Between Four-Week Body Weight Expressed in Two Dietary Environments (for the High and Low Lines) . . . . . . . . . . . Number of Chicks Hatched, Number Surviving at Four Weeks of Age, Mortality Percentage and Survival Sex Ratio by Rearing Environment, Generation and Lines . . . . . . . . . . . . Expected Phenotypic Correlations by Various Methods for the Selected and Correlated Traits Within Line 0 O O O O I O O O O O O 0 Least Squares Estimates for Main Factors and First Order Interaction Effects with S.E. for the Different Sources of Variation for the Subclass Means for Four—Week Body Weight of Fourth Generation Chicks Reared in two Dietary Environments . . . . . . . . . . . . . . . . Adjusted Subclass Progeny Means (by Lines and Selection of Parents) in Two Rearing Environ- ments (Fourth Generation) . . . . . . . . . . Analysis of Variance of Four—Week Body Size for Generation 4 Whose Parents were Selected From Environments Where Birds Received Rations Containing Either Raw or Cooked Soybean . . . vii Page 81 84 87 91 94 95 99 102 105 106 108 u _ . a L r; o . .U I. e t u .uw VA \. (A J . .wx HP. _.. t .— e .u n? V. ~1 .d up. d U... h... an.» .4 .Cu sh .. Tu 8 Au — “Lu Y M In INC CIV- 9i¢ «IL. «(J- 5. u «I/s «l;- fil‘ nl'l ll 1 .ll Table Page 13. Selection Differential, Response and Realized Heritability for Four-Week Body Weight for Line by Generation . . . . . . . . . . . . . . . . . . 81 14. Mean Percentage of Fertility, Hatchability and Hatching Abnormalities (Adjusted for Line Effect) as Correlated Response to Selection of Body Size at Four Weeks of Age for the High and Low Lines Over Generations . . . . . . . . . . 84 15. ANOVA Between Rearing Environments Within Line for 4—Week Body Weight by Generations . . . . . . . . . 87 16. Estimated Components for Computation of Genetic Correlation by Various Methods . . . . . . . . . . . . 91 17. Sire-Progeny Unadjusted Means for Body Weight in Grams at Four Weeks of Age by Lines, Generations and Rearing Environment . . . . . . . . . . . . . . . . 94 18. Estimated Pooled Genetic Correlation over Generations Between Four-Week Body Weight Expressed in Two Dietary Environments (for the High and Low Lines) . . . . . . . . . . . . . . . . 95 19. Number of Chicks Hatched, Number Surviving at Four Weeks of Age, Mortality Percentage and Survival Sex Ratio by Rearing Environment, Generation and Lines . . . . . . . . . . . . . . . . . 99 20. Expected Phenotypic Correlations by Various Methods for the Selected and Correlated Traits Within Line . . . . . . . . . . . . . . . . . . 102 21. Least Squares Estimates for Main Factors and First Order Interaction Effects with S.E. for the Different Sources of Variation for the Subclass Means for Four—Week Body Weight of Fourth Generation Chicks Reared in two Dietary Environments . . . . . . . . . . . . . . . . . . . . . 105 22. Adjusted Subclass Progeny Means (by Lines and Selection of Parents) in Two Rearing Environ- ments (Fourth Generation) . . . . . . . . . . . . . . 106 23. Analysis of Variance of Four-Week Body Size for Generation 4 Whose Parents were Selected From Environments Where Birds Received Rations Containing Either Raw or Cooked Soybean . . . . . . . 108 vii . r . . . . I I . .I .. . . . o. C . . Va . . . . . ~ v n .. . u a a . t . . . . . e . Y .9 ‘ Iq .. s IA : L vet . r: —-.~ C D ’0‘ an. ab .0 I. IL r». e a Mn .0. _ a .P. re. A. . ..4. P... .v. a»; tfiu Ct nit h r e ..b 4 a u . . .. . n a ntt A G Um g H . T. . . n a (v We D 0 Co e “d NJ. . E 3 l u a v A at P t 5 CA L e mi u -c r . I. ‘ .xJI U l . . . I «6 Al‘ 3 .. V I I .. . . .3 ‘6 o u FIR W, film» I If [(ll‘llll‘l . ll Figure LIST OF FIGURES A Path Coefficient Diagram Showing the Correlations Involved When a Sire is Mated to Several Cows to Produce Offsprings to be Reared in Nutritional Environments Where the Ration had raw (R) or Cooked (C) Soybeans . . . . . . . . . . . . . . . . Growth Curves from Hatching to Four Weeks of Age for Offspring of the High and Low Line from the First Generation to Fourth Generation Receiving Growing Rations Containing Raw or Cooked Soybean . Mean Response for Selected and Correlated Traits (Four-Week Body Weights) for the High and Low Line Offspring Fed Diets Containing Raw or Cooked Soybean . . . . . . . . . . . . . . . . . . The Effect of Genotype-Environment Interaction as a Deviation from Overall Constant by Line and Environment Within a Generation Basis . . . . . . . Independent Effect of Genetic and Environmental Forces in Determining the Net Change in Four- Week Body Weight per Generation for the High and Low Line Offspring Receiving a Diet Containing Raw Soybean . . . . . . . . . . . . . . The Effect of Selection Over Four Generations as Divergence Between the High and Low Line Offspring Receiving Ration Containing Raw Soybean . . . . . . . . . . . . . . . . . . . . . . The Fitted Quadratic and Linear Lines for the High and Low Line Mean 4-Week Body Weights (gms) of Offspring Fed Raw Soybean Ration Over Four Generations . . . . . . . . . . . . . . . . . . . Observed and Expected Response to Direct Selection for Four-Week Body Weight in the High and Low Line Offspring Receiving the Raw Soybean Ration . viii Page 38 54 63 64 68 69 70 73 Figure 10. Figure Page 9. Population Parameter Heritability Estimate of 4—Week Body Weight for Line by Rearing Condition and Generation . . . . . . . . . . . . . . . 79 10. Correlated Response of Fertility and Hatchability Percentages to the Selection of 4-Week Body Weight for the High and Low Line Offspring . . . . . . 85 11. Estimated Genetic Correlation Between Selected and Unselected Traits Within Generation by Lines (Robertson's Method) . . . . . . . . . . . . . . 93 12. The Effect of Sex by Rearing Environment as an Interaction Deviation from Overall Constant Within a Generation Basis . . . . . . . . . . . . . . . 98 13. Geometric Presentation of the Performances of the High and Low Line Progeny Sired by Third Generation Parents Being Reared on Ration Containing Raw or Cooked Soybean . . . . . . . . . . . 107 ix raw .5 .. . .8 r . .2 e .1 u 9; a.“ .l INTRODUCTION At present, for rapid growing broilers, the high-energy ration formulated by the nutritionists invariably consists of a major fraction of soybean meal because of its rich source of quality protein and amino acid balance. Furthermore, all these rations contain processed rather than raw soybeans. The cost of processing soybean to extract oil for human consumption was, in the past, proportional to supply and demand and therefore was economical, but due to the rapid automation of the livestock and poultry industry, the demand for soybean meal in bulk may outweigh the demand for soybean oil. So the oil extraction process in the future may not be economical. However, it may not even be neces- sary if the soybean as such could be utilized by the livestock as efficiently as the processed forms to meet the growing demand for animal protein by a growing population. A great deal of work, resulting in many research papers, has been done on this particular aspect of raw soybean utilization by all kinds of livestock. One, and the only one, conclusion from these re- search works is the "growth depression" of the individuals when fed with diets having raw soybean regardless of supplementing the ration with additives to augment the growth or correctives to counteract the depression. Fortunately, the "growth-depressant factors" as they are called, are heat labile. Curiously enough, the chicken at certain stages of 1 2 age (2—4 weeks) are most vulnerable to the growth depression of raw soybean in the ration but grow and produce eggs equally well at adult age (20-22 weeks) when receiving rations containing either raw or cooked soybeans. From a review of literature regarding the growth depressant factors present in raw soybean, it was observed that these depressant factors in some way or other are related to the metabolic efficiency with regard to growth rate at specific ages. Since the metabolic efficiency of a phenotype is directly or indirectly concerned with the genotype which interacts with its environment to be expressed as a phenotype, the selection of such a phenotype growing better or poor when receiving a ration containing raw soybean may indirectly estimate its metabolic efficiency or inefficiency. Even if a phenotype grows better than others when receiving a ration containing raw soybean, it does not necessarily mean that it has a superior metabolic efficiency at the genetic level as compared to that of other phenotypes in the group. If this superior growing ability cannot be perpetuated it is only a phenotypic adaptation to adjust the individual in question with the environment temporarily for its survival and so it is purely a non-genetic event. Therefore, to estimate this transmission of the superior growing ability from one generation to the next by the individuals fed with raw soybean ration, a bi-directional selection program utilizing chickens of a rapid growing stock was thought to be worthwhile to pursue. OBJECTIVES To establish two lines of chickens on the basis of 4-week body weight, one selected for large body size and the other for small body size, when receiving a diet containing raw soybean. To estimate genetic parameters such as heritability and genetic correlation between the performances of chickens in stressed and normal nutritional environments. (Knowing these parameters would facilitate progress in future selection experiments to improve body size of chickens.) To understand the genotype—environment interaction in order to pre- dict the growth response in normal situations where selection is made from a stressed environment. To determine whether or not a sex difference in response to utilizing raw soybean exists. To estimate reproductive fitness, i.e. fertility and hatchability, when the two selected lines are subjected to a stressed environment. REVIEW OF LITERATURE Growth Depressing Factors in Raw Soybean In 1917, Osborne and Mendel pointed out the growth promoting effects of cooked versus raw soybeans. Raw soybean contains several substances which impair growth, cause pancreatic enlargement (not in all species), retard the intestinal and pancreatic proteolysis by their respective enzymes and may interfere with intestinal absorption as well as the metabolic functions of different body organs. Fortunately, these growth depressing substances are heat labile. In addition to the above subjective effects, certain indirect factors, like species, age, etc., also influence the raw soybean effects on the animals, e.g. pigs do not develop pancreatic enlargement. Trypsin and Chymotrypsin Inhibitors Kunitz (1946) first crystallized a soybean inhibitor (CSBTI) which contains mostly tryptophan. The other inhibitors of the raw A soybean that have been isolated are F AA, and 1.93. Most 1’ F3’ 1’ of these trypsin inhibitors inhibit with a varying degree the Chymotrypsin (Birk, 1961; Rackis and Anderson, 1964). The order of soy trypsin inhibitor potency of the soy fractions, characterized to date is CSBTI = A = AA = 1.93 > F > F3 (Frattali and Steiner, 1968). Simi- l l larly, the Chymotrypsin inhibitor potency order is 1.98 > AA > A1 > CSBTI > Fl > F3. However, the biochemical reaction of these inhibitors 4 5 so far known, has been described as forming stoichiometrically an inseparable complex with trypsin. Sealock and Laskowski (1968) showed that this complex was due to a close cystine disulphide bridge across the arginine residue which did not allow the trypsin inhibitor molecule to open and release the trypsin once the arginine residue had been hydrolyzed. The AA inhibitor is non-competitive but the CSBTI is competitive to proteolytic and estrolytic activities of trypsin and Chymotrypsin. The depression of growth by adding CSBTI to chick (Garlich and Nesheim, 1966; Gartler gt 31., 1967) and rat (Gartler_e£ial., 1967; Haines and Lyman, 1961) diets was not as great as that caused by raw soybean meal. Further, the crude crystalline trypsin when added to raw soybean meal diet for chicks (Brambilla g£_al,, 1959) and rats (Borchers, 1961) failed to counteract growth depressing effects of raw soybeans. Feeding amino acid instead of intact protein did not inter- fere with the intestinal enzymatic digestion. Therefore, the growth depression from feeding raw soybeans seems to be something other than the inhibition of intestinal enzymatic activity and since only crude trypsin inhibitor preparations were used in all studies, the possibility exists that the causative factor might not be the trypsin inhibitor only. Hemagglutinins Liener and Pallansch (1952) purified a haemagglutin of 96,000 mol. weight containing 6-10Z glucosamine from defatted soybean flakes. Lis ‘gg_§1. (1966) found four hemagglutinins in soybean oil meal, which could be separable in DEAR-cellulose columns. However, all these are III'II'IIIIIIIJ‘I III I|IIIIIIII III IIl-III‘IIII‘.I|III 6 glycoprotein containing mannose and glucosamine. Although intraperi- toneal injections of hemagglutinin preparations were lethal to young rats (Liener, 1951), this is not sufficient evidence to account for the growth depression attributed to hemagglutinins because of the following reasons: 1. Hemagglutinin is already inactivated by peptic digestion, even when as few as 12 percent of peptide bonds are split (Birk, 1961); therefore, it should be either completely or almost completely inactivated before entering into the small intestine. 2. Even if hemagglutinin does survive the gastric digestion, an intact protein of 96,000 mol. weight would not likely be absorbed from the gut. Saponins At least 5 different saponins have been isolated which exhibit varying degrees of hemolytic and foam—producing activity (Willner g£_al., 1964). It is seen that high levels of soybean saponins inhibit pro- teolytic activity of trypsin and Chymotrypsin. However, Birk gt_al. (1963) showed that the hemolytic activity was unaffected by the heat treatment required to produce optimum nutritive value of soybean meal, indicating, therefore, that the hemolytic property of saponin has little or no influence on soybean nutritive value. Pancreatic Enlargement Pancreatic enlargement is a concomittant effect to growth depression, especially in chicks and rats fed a raw soybean ration. This is presumably a compensatory mechanism of the pancreas to reduced 7 proteolytic digestion in the intestine caused by the trypsin inhibitors. Histological examination showed pancreatic hypertrophy associated with accumulated zymogen granules in the acinar cells (Booth gt 31., 1960; Saxena_g£-§1., 1963). An increased secretion of pancreatic enzymes also accompanied pancreatic enlargement (Lyman, 1957). Further, it was shown by Saxena 25 a1. (1963) that the pancreatic enlargement was not an irreversible process, because alternating the raw and cooked soybean ration every few days increased or decreased, respectively, the pancreatic activity thus supporting the compensatory hypothesis of the pancreatic response in feeding of raw soybeans. Loss of Endogenous Nitrogen Pancreatic enlargement and increased enzymatic production by animals on raw soybean meal suggests that increased removal of endoge— nous nitrogen might result (Lyman and Lepkovsky, 1957). Lepkovsky 35 31. (1965) observed a higher proteolytic activity in feces of rats fed raw soybean meal than in that of rats fed cooked soybean. This supports the hypothesis that, over a long period, the excess nitrogen loss from the body might need an increased dietary protein which, therefore, accounts for some of the growth depressing properties of raw soybeans. Metabolic Effects Conversion of methionine to cystine was shown to be influenced by feeding raw soybean or its growth depressing factors (Gorrill and Thomas, 1967). Khayambashi and Lyman (1966) observed higher cystine content in the intestine of rats fed raw soybean as compared to that of cooked soybean fed rats. Kwong and Barnes (1963) showed that 8 feeding of raw soybeans increased the expiration of C140 by rats when 2 they were injected intraperitoneally with D-L—methionine-Z-ClA. This increased metabolism of methionine as measured by C02 loss did not occur in rats fed RSBM (raw soybean meal) diets supplemented with cystine. In later experiments, using L-methionine-S35 these researchers showed that methionine was rapidly converted to cystine in the pancreas of rats to which crystallized soybean trypsin inhibitor was being administered through stomach tubes. Impaired Intestinal Digestion and Absorption Decreased protein digestion and amino acid availability are mostly accepted as the possible major factors in explaining the growth depression caused by raw soybean. Since methionine is the most limiting amino acid in soy protein for rats (Kwong gpflal., 1962), chicks (Snetsinger and Scott, 1958) and poults (Linrode g; 31., 1961), the decreased availability of sulphur amino acid or an increased total requirement might cause a growth depression in animals on diets con— taining raw soybean. Digestibility studies indicated that raw soybean meal contains a fraction which became digestible only after cooking (Nitzan, 1965). However, there is no agreement as to whether the absorption of sulphur amino acids is influenced by heating soybean meal. Influence of Other Amino Acids An improvement in growth of rats fed a raw soybean ration supplemented with tyrosine, valine, threonine and methionine has been reported (Mickelson and Yang, 1966). 9 Very recently Schingoethe (1968) isolated a small molecular weight growth inhibitor from soybean trypsin inhibitor by ion-exclusion chromatography on a sephadex G-SO column. This growth inhibitor de- creased the weight gain and feed efficiency of mice without causing pancreatic enlargement. Heritability Heritability is the key parameter for a given population, if the genetic improvement of the population is desired. In the context of one environment, the genetic control of a character at the level of statistical statement is only understood when it is quantified as heritability. Heritability estimates in the narrow or broad sense (Lush, 1948) specify the context of additive and non-additive genetic factors in— volved. In the narrow sense, a heritability estimate is the fraction of the additive genetic variance accounted from the total phenotypic variance and in the broad sense it is the ratio between the additive and non-additive genetic variance with the total phenotypic variance. Body weight, being the total phenotypic expression of many com- plex underlying physiological mechanisms, is influenced by multiple genetic factors (Asmundson and Lerner, 1933), additive effects and non-additive effects (dominance, epistasis). Sex-linked gene effects and maternal effects also influence the body weight. In view of the non-additive genetic effects on body weight, several workers attempted to determine their relative importance in chickens. Martin_e£_§1. (1953), Brunson 33 a1. (1956), Goodman 25-31. (1957), Goodman g£_al. (1960) and Kan $5.31. (1959) found that non- additive genetic effects were small; whereas, Yao (1959, 1961) 10 observed a significant dominance effect in certain crosses and incross- breds. Maternal effect influencing the differential body weight at 10 weeks of age was pointed out by Yao (1961). Brunson 25.31. (1956) and Thomas g£.al. (1958) reported sex-linked gene effects in the inheritance of body weight. From a bi-directional selection experiment for body weight at 8 weeks of age, Siegel (1962) discussed the insig- nificant role of natural selection and epistatic deviation. He found the additive genetic variance to be 30 percent of the total variance. Further, he presented a good summary, based on 176 heritability esti- mates for 6 to 12 weeks of age, computed by different methods in dif- ferent populations by different workers, where the range of this esti— mate was from -.20 to 1.00 with the median at .41 having an interquartile range of .29 to .54. This, therefore, indicates that the body weight at different ages is a moderate to highly heritable trait and mass selection could bring a genetic change for this trait in a population. For all practical purposes the half sib correlation from sire variance component is the best estimate of heritability and is unbiased for any dominance deviation and is regarded as heritability in the narrow sense. Since full sibs are correlated in this dominance deviation to an extent of 1/4, to a certain extent for epistatic deviation (= 3/16) and all permanent environmental maternal effects, the heritability estimate obtained from dam variance component may approach the upper limit and is generally heritability in the broad sense, the concept introduced by Lush (1948) 2 h2 = o p - V[E] . 2 0 P 11 The other methods of estimating heritability, e.g., regression of offspring on parents, realized heritability in selection experiments, are also practiced to estimate this parameter. The realized heri- tability of a selected trait may be obtained either by (l) dividing the cumulative response by the cumulative selection differential or (2) by regressing cumulative response on the cumulative selection dif- ferential (Falconer, 1960). The realized heritability from a two-way selection could be a more reliable estimate than that from nested variance analysis as it is based on the true response over a period of generations. This minimizes the bias due to common environment effects, epistatic and dominance deviations. Ideta and Siegel (1966) reported the realized heritability for body weight of chickens at 8 weeks of age to be .31 for males and .28 for females by the first method, i.e., dividing the cumulative response by the selection differential. By the second method, i.e., regressing the cumulative response on cumulative selection differential, they obtained estimates of .35 i .05 and .31 11.03 for males and females, respectively. The realized heritability for 9 months body weight in a bi-directional selection experiment by Nordskog and Festing (1962) was .41 for the high weight line and .85 for the low weight line. "Correlated Heritability? of Unselected Trait The term correlated heritability of a trait has been proposed by Carte and Siegel (1968) instead of the term realized heritability of an unselected trait as used by Ideta and Siegel (1966). They pointed out that the response for an unselected trait may be directly measured on the selection differential. Magee (1965) in discussing II.llII.[.I[ III! [III 12 the response to selection did not approve the use of selection dif- ferential of an unselected trait but suggested that it be called a secondary selection differential. He pointed out that the term selec- tion differential should be reserved for a trait directly under mass selection in the context of the terminology used by Lush (1945). This is because the changes in the correlated trait are not usually equal to the heritability of correlated traits times selection differential (‘13—‘17). When the selected and unselected traits are correlated and distributed normally, the expected secondary selection differential for the unselected trait is equal to the selection differential of the selected trait multiplied by the phenotypic regression coefficient of the correlated trait on the selected trait (see Carte and Siegel, 1968 for details). Heritability estimates for poults of different ages (2, 4, 6, 8, 16, 24 and 25 weeks) were calculated, using the analysis of variance technique, by Bumgardner and Shaffner (1954). The estimates from sire (48) P+D+S had a range from .003 to .464, the lowest component of variance being at 4 weeks of age and the highest at either 8 or 16 weeks of age. They concluded that despite the maternal influence at early ages in both sexes, mass selection based on individual phenotype would be successful in improving the breeding ability for growth of the poults because of a moderate to high heritability (.30 to .48) for this factor for ages beyond 8 weeks. In Fayoumi, Amer (1965) reported the heritability estimates of body weights at hatching time, 4 and 8 weeks of age to be .36, .54 and 13 .76, respectively, on the basis of both parent's contribution of addi- tive genetic variance to their Offsprings [Zfiégggl . Genotype x Environment Interaction The existence of genotype by environment interaction is under- stood by the fact that the best genotype in one environment is not necessarily the best in another environment. This is probably due to the physiological mechanisms involved in expression of the same trait of a given genotype in different ways in different environments. Therefore, the accuracy for the prediction of response in a selection experiment for an unselected trait is restricted as a result of shifting of the phenotypic ranking of a series of genotypes from one environment to another unless a prior genotype x environment effect is known. Falconer (1953, 1960) introduced the concept of genetic correlation between two traits (Hazel, 1943) which could be extended to the genetic correlation between the phenotypes of the same genotype expressed in two different environments. He proposed that this kind of genetic correlation could be the appropriate measure of genotype-environment interaction. Robertson (1959) suggested that if the genetic correla- tion is not significantly deviated from 1 or .8, then genotype by environment may be considered negligible due to the standard error of .2 which should not be taken as a high deviation in a parameter such as this, which is very sensitive to sampling errors. Dickerson (1962), presenting an analysis by two-way classification, showed how to measure the genotype-environment interaction by estimating the genetic correla— tion. He pointed out that genetic change in performance under one environment (AGj) for response to selection based on performance under 14 another environment (AGi) is directly proportional to the genetic cor- relation of performance under the two environments or inversely pro- portional to the importance of genetic-environmental interactions which causes shifts in genetic ranking of the phenotypes between environ- ments. Yamada (1962), attempting to compare the different methods of estimating genetic correlation using fixed and random models, proved that regardless of statistical model the genotype (group) by environ- ment interaction variance is equal to the average of the two between- group variances under each environment minus the between group variance. Mather & Jones (1958) elaborated the theory of genotype by environment interaction and by means of three parameters, da’ e and g, i.e., average genetic, environment and genotype x environment effects, respectively. They discussed the six possible phenotypic relationships between two genotypes (AA, aa) in two environments (x, y) depending upon the magnitudes of these parameters, da’ e, and g. Where the dif- ferences between genotypes are smaller as compared to the differences between the environments (most usual circumstances, generally), the da and g must be smaller than e and in that case the four distinct phenotypes will be such that the relationship between these three parameters is e > da > g. They further described the model to estimate the genotype x environment interaction in a polygenic trait. The familiar demonstrations of genetic-environmental interaction, such as genetic differences in disease resistance under variable exposure (Hutt_ep.§1., 1945), in sensitivity to nutritional deficiencies (Lamoreux and Hutt, 1948), and in responses to growth under high and 15 low plane of nutrition (Falconer, 1959), are unique in their respective areas. Whereas reports regarding genotype x environment interactions in egg production stocks are numerous in the literature, those for broiler-production stocks are very limited. With regard to genetic difference in performance of floor vs cage housed birds, the inter- actions concerning the shift of sire and dam progenies are not clear- cut since Gowe (1956), in seven Leghorn strains, demonstrated signifi- cant strain by environment (housing) interaction for March body weight (rG = 0.30) and survivor egg production (rG = 0.25) with regard to cage vs floor housing; whereas Lowry 25 a1. (1956) did not observe any such interactions. However, Hill and Nordskog (1956), with more samples of genetic stocks (13 stocks) and farm environment locations (11 farms) showed highly significant stock x farm interaction for adult mortality (rG = .03), egg production and income index (rG = 0.6-0.7), sexual maturity and egg size (rG = 0.75 - 0.80), body weight (17 months) (rG = .85) and specific gravity of eggs (rG = .60). Similarly, from California Random Sample tests results for a three year period, signifi- cant stock x year interactions for many traits, such as egg production (hen-day, rG = .70, hen-housed rG = .69), egg size (rG = .89), albumen score (rG = .74), income index (rG = .68), etc., were reported by Dickerson and Abplanalp (1956) and were in agreement with the findings of Hill and Nordskog (1956). After using two rearing environments (restricted and full feeding), Gowe gpflgl. (1962) pointed out the importance of genotype by environment interaction. In all traits they considered on within year analysis, they found a high genetic correlation (rG = 1.0) and therefore 16 questioned the genetic-environment interaction. However, Abbott and Asmundson (1962) could demonstrate the efficiency in responses to selection under a severe stress environment. Their studies were in- volved with the use of a mutation (scaleless birds) which through its effect on feathering placed the bearers under severe hypothermic stress. They observed a significant rise in viability when the birds were selected for this trait from the stress environment as compared to viability of normal birds. Further, appreciable improvements in growth and feed efficiency were also observed in most of the lines they studied and their overall conclusion was that the most fit scaleless bird belonging to a given line was a large sized female (sex influenced) heterozygote with efficient feed conversions, having more down feathers than average for its line and produced from a crossing between a normal dam with a scaleless sire. The only weak point in their studies was that they did not analyze their data in such a way as to demonstrate the genotype x environment interaction. Genetic Correlation The usefulness of genetic correlation is three—fold: 1. To account for the genetic cause of phenotypic correlation between two metric traits. 2. To predict a correlated response on a trait, caused by selection on another trait. 3. To find out the genetic prOperty of a metric trait and its way of association with fitness in view of natural selection. The phenotypic correlation between two metric traits is comprised of two components (genetic and environmental) and is expressed as r = P e xy h x hy rA + xeyre (see Falconer, 1960 for details). 17 When there is no environmental correlation (re), the phenotypic correlation (rp) is exclusively determined by genetic correlation (rA) and the heritability of the traits (x, y). This happens when the same trait is measured in two different environments in different individuals belonging to the same group or pedigree and is relevant to our problem. The reader is referred to the sections on genotype by environ- ment interaction and correlated response for a detailed review of literature. Correlated Response A correlated response denotes the associated changes on an unselected trait when selection is made for a given characteristic. The theory of correlated response has been fully discussed by Lerner (1958) and Falconer (1960). The theoretical treatment of genetic rela- tionships between two or more characters leads to the conclusion that the correlated response depends primarily on the genetic correlation between the two characters due to pleiotropy and linkage. Hazel (1943) has shown the implication of genetic correlation in selection experi- ments when two or more characters are chosen for simultaneous improve- ment. Reeve and Robertson (1953) and Clayton e£_al. (1957) showed a confused picture of correlated response for unselected trait. The latter workers while selecting for increased and decreased sternital bristle number observed asymmetry in response of sternopleural bristle number and therefore concluded that genetic drift played an important role, in the presence of a low genetic correlation, to bring the correlated response. Falconer (1953) who used mice in two-way selec- tion studies (for body size in one pair of line and for tail length in 't‘llltllllllIr‘ (III 18 other pair) showed complete agreement between the theoretical implica- tion of genetic correlation and observed true correlated response. If the phenotypic expressions of the correlated traits are dependent on alleles with a general effect on metabolic efficiency or on the same hormone, the correlated response may be due to pleiotropy (Lerner, 1958). On the other hand, linkage causing correlated response between traits may initially be present in the population or may be the result of selection. In the latter event, the polygenic blocks are integrated by selection while locking few alleles other than the selected ones and therefore the correlated response is observed for a few generations until the unselected alleles are unlocked by crossing over. The general fitness characters such as fertility, livability and mortality in animals, including hatchability in birds, unless selected, are observed as correlated response in selection experiments for metric traits. According to Falconer (1960) the change in the metric traits by artificial selection brings a reduction in fitness characters and this is attributed as a correlated response to maintain the "genetic homeostatis". Nordskog (1962) concluded from his selection experiment in White Leghorn and Fayoumi with regard to correlated response in fer- tility and hatchability in selected breeders that there was a marked drop in fertility in large bodied Leghorns and Fayoumi. He did not observe any correlated response in hatchability. However, he pointed out that large body-size or large egg size are unfavorable to high hatchability whereas the reverse selection for these traits favors high hatchability. II.|IIII.IIII((II‘[I l9 Lerner and Gunns (1952), in their studies of hatchability (a reproductive fitness in general) in relation to egg weight, pointed out that in flocks where some artificial selection for large egg size is practiced, maximum hatchability is obtained in eggs somewhat below the mean egg size, i.e., smaller egg size favors high hatchability. Nordskog (1962) observed the positive correlated response in egg weight when selection was made on nine months high and low body weight re- spectively in Leghorn and Fayoumi and therefore, subsequently the hatchability as a correlated response was better in the low body weight eline than in the high body weight line. In discussing the five factors contributing toward the asymmetry in selection response, Falconer (1960) was unable to support any single factor as being responsible for this asymmetry. Bohren 35.31. (1966) discussed the asymmetry of correlated re- sponses in two traits or of the same trait in two environments. They used several models of gene effects and gene frequencies and observed that asymmetry of the genetic covariance and consequently of the cor- related response resulted when the relative changes in the gene fre- quency at the loci contributing positively or negatively to the co- variance depended on the trait selected. They concluded that the most frequent contribution to asymmetry in practice is due to the loci con- tributing negatively to covariance and the gene frequencies being other than 0.5. They concluded that an accurate prediction of a correlated response over several generations is not possible in the absence of knowledge regarding the composition of genetic correlation and its magnitude. 20 Sex Linkage and Sex Effects In chickens, very few studies regarding the effect of sex— linkage on quantitative traits have, so far, been made. Since prob- ably only a few sex-linked gene mutations would show any qualitative effect and because the sex chromosomes are a very small proportion of the total number of chromosomes (Krishan and Shaffner [1966] reported that there are 39 pairs of chromosomes in aves), sex—linkage for quantitative traits was somewhat ignored until Newcomer (1957) presented evidence for only six pairs of chromosomes (the rest of them being con- sidered chromosomoids; that is, reserved materials) in the domestic fowl. As a result of Newcomer's research, the importance of sex- linkage in metric traits has drawn the attention of many workers. Thomas e£_al, (1958), while reporting the heritability estimates of different traits, such as feed efficiency, gain, feed consumption and body weight at 4, 6, 8 and 10 weeks of age for broiler type New Hampshire strains, discussed the possibility of sex-linkage as a cause for the divergent estimates of heritability in body weights, calculated from sire component for male and female in order of ages. In the study of heritability estimates for egg weight, shape and number in a New Hampshire flock, Hicks (1958) pointed out a sex-linked gene effect on egg weight due to the higher values of sire component than that of dam in five out of six estimates. The other relevant studies of sex-linkage in different traits in chickens deserve citation in brief——percent batch of all eggs and laying mortality (Goto and Nordskog, 1959), body weight at 10 weeks of age (Brunson_e£.al., 1956), adult mortality (Nordskog and Phillips, 21 1960), mature body weight (Jerome g£_§l,, 1956), and egg shape index and egg weight (Goodman and Jaap, 1961). In white Holland turkeys at 16 and 24 weeks of age, McCartney (1955) who estimated the heritability by using variance component analysis and offspring-parent regression methods, pointed out the influence of sire on daughter and dam on son in contributing the addi- tive genetic variance due to the sex-linked genes. STATISTICAL PROCEDURE Because of unequal frequencies of observations within subclasses, the analysis of variance by fitting least squares constants to the various variables considered in the model was the best approach to estimate the effects of the main factors and their interaction effects. The main factors in this experiment were two genotypes with regard to group average for selection for high and low body size and two rearing nutritional environments (raw soybean and cooked soybean ration). Due to the influence of sex on body weight regardless of selection or rearing environment, it was imperative to consider this as another factor contributing its effect to the mean performance. Further, the nested sires within line and nested dams within sires within line, being other random factors influencing the average performance, were included in the model. Thus, the model had three fixed factors (rearing environment, line and sex) and two random factors (sires and dams) with their re- spective first and second order interactions. However, to make the appropriate tests (Table 1) for the significance of various main and interaction mean squares (except the line and sires), the random inter- action source of variations due to sires and dams with all other fixed factors were included in the error term. The programming for the raw data was made in 3600 Fortran com- puter language according to the stat description Series No. 7 and 18 22 23 x n AHIEVAHIHv\NthA m w HHU AHIEVAHIHV xmm x mafia N N x N H AHIEVAH-pv\Ntzm N N o o AHIEVAHICV xmm N N x ucmEcouw>cm wcflummm h a o AHIHVAHIuv\NanmN N N o AHIHVAHICV mesa N N x ucmEcouH>cm wcflummm HIE\Nxz N mo H I E xmm p N o m I p ocHH\mmuHm\mEmm Nmo 00 + p mo H I m ocHH\moHHm Huh a m m e N HIo\NnA m u + N o o + u H I a mafia HIH\NNM N No H I u unmecoufi>co wcwumom Amzvm mu mmousom wsflumop somzuob pom mmcfia coozuon <>oz< HomV mumsvm Gama vmuoomxm ¢>oz< .Amuamaaouw>cm .H maan 24 H I z AHIaVAHIHVAHIuv\anszqm otw.¥ NW“ NN-r-I NH U + .O .O UIAHBIA +.....G AHIEVAHIHVAHIAV HmuOH uonum xom x mafia x unmanouw>co wnfiumom 25 writen by Rubel e; 31. (1967), Department of Computer Science, Michigan State University. The model used for the analysis of variance within generation was as follows: = u + R. + L + RL + + RM. + LM. + RLM.. + s . + D + Error y 1 J 13 ”k 1k 3k 13k pa qu Error = e. + SR . + DR . + ......etc. ( ljkpql p31 qui ) where: u = Overall mean R1 = the effect of the i?? rearing environment when i = l, 2 L. = The effect of the 3?? line when j . 1, 2 J RLij = the interaction effect of the 1?? rearing environment with j?? line. Mk = the effect of kE§ sex when k = l, 2 RMik = the interaction effect of if? rearing environment with k?? sex LM - the interaction effect of j?? line with kg? sex jk RLMijk = the second order interaction effect of if? rearing environment with j?? line and k?? sex th th Spj = the random p.. sire effect with j.. line when p = l ------- s quj = the random qf? dam effect within pEP sire within jg? line Error = random error plus the random interaction effect of p?? sire and qf? dam with k?? sex and if? rearing environment Restrictions imposed: ER = 2L = ZS = ZRM. = ELM. = ZRL. = i j k ik 1k jk jk ij 1j .z RLM = 23 = 2D = o . - i'k 1jk J pj pj qu The expected mean square with degrees of freedom are presented in Table 1. The coefficient of variance components for the expected 26 mean squares were obtained for the nested portion of the model according to the formula by Kempthrone (1957). The coefficients of interaction sum of squares for fixed factors (C9, C ) were obtained by 10’ 011’ C12 averaging the number of observations per treatment combinations within cells. These estimates are not exact, but approximate. The genotype-environment interaction effects, which imply the real differences between the group of relatives, are not the same in two environments and may arise due to two biologically different entities: (l) the between group variance components may be different; (2) the true ranking of the groups may be different. Since the second cause has more implication in consideration to genetic correlation, it was decided to run a two-way analysis of variance within line where the model considered the rearing environment as the fixed factor and sires and dams as being random factors. The interesting point in this analysis lies in the estimation of the sires by rearing environment or dam by environment interaction variance com- ponents. Robertson's (1959) concept of genotype-environment inter- action considered in a fixed model for estimation of genetic correla- tion has been further extended by Yamada (1962) to mixed and random models. Thus, primarily the concept of Robertson (1959) with the help of mixed model in the procedure of Yamada (1962) has been used in this experiment while analyzing the data within line. Therefore, in the mixed model in our experiment, the following were the factors: Y = p + R + S + D , + RS + Error 1 j k3 ij (Error = eijkl + DRkji) 27 Where: u = overall mean R1 = the effect of 1?? rearing environment when i = l, 2 Sj = the effect of j?? sire when j = 1 ----- S ij . the effect of kFI dam within jf? sire when k = l ----- D R = the interaction effect of if? rearing environment with j?? sire 11 Error 8 the random error plus the interaction effect of iFI rearing environment with kEI dam within jg? sire Restriction imposed: :Ri = §Sj = kaj =i§RSij = 0 However, in the complete model, the dam-environment interaction should have been considered as a separate source of variation rather than including this into the error term. This was not possible in our data because of the difficulties encountered in inverting the matrix due to discrepancies in orders and ranks (singular matrix) when this piece of variation (dam-environment interaction) was considered as an individual source of variation in the model. Instead of deleting some of the dams (according to Stat description series LSDEL or LSADD) it was rather convenient to put the source of this dam—environment variance into the error term. With regard to the full model, the proper F-test for sire-environment mean square should be on dam-environment inter— action mean square for the test of significance. If the dam- environment interaction source of variation is included in the error term (as in this analysis) the F—test for significance of sire- environment interaction would take the error term in the denominator. Therefore, there is the possibility of bias either upward or downward 28 depending upon the magnitude of the difference in the coefficients of variance component due to dam-environment source of variation included in the error term and the same appearing within the sire—environment interaction variance. If the dam-environment interaction variance is assumed to be zero, there is no bias in the testing procedure at all. However this was necessary to calculate the genetic correlation by two-way analysis of variance and then to compare the same estimate computed by variance and covariance technique as suggested by Magee (1969). The expected mean square, degrees of freedom and the K values are presented in Table 2. The coefficient of variance component for sire-environment interaction (kl) was obtained by the formula derived from Henderson (1953) Method 1: A kI = [N - Z Z nij.2/ni.. - Z X nij.2/n.j. + £2 nij.2/N /df i j i j i j where: i = rearing environment j = sire n = number of observations per treatment combination N = total number of observations df = degrees of freedom for sire by environment interaction Genetic Correlation A. By Two-Way ANOVA — The genetic correlation estimated from the formula by using the mixed model in the outline described by Yamada (1962) is as follows: Table 2. within line). 29 Expected mean square (ANOVA between rearing environment Sources df E(MS) 2 2 2 2 Rearing environment r - 1 o + k 0 + k X Ri /r-l w l R 2 1=l 2 2 2 Sires s - 1 0w + k3od + k4oS . 2 2 Dams/Sires d - s o + k 0 w 5 d Sire x rearing 2 environment (s-1)(r-l) ow + k6OsR2 2 Error n... + (s—d-sr) oW Total N — 1 30 o *2 2 r = 812 = o s - 1/2 0 sR G12 051.032 023 + 1/2 02 SR - 1/2 6 (bi) Where: 028 = estimated variance component of sires between rearing environments 028R = the estimated variance component due to sire by rearing environ- ment interaction 6 (bi) = the estimated squared difference between standard deviations of the variance component due to sires within respective rearing environments B. By Variance and Covariance Technique - Falconer (1960) intro- duced the concept of considering one trait as two when it is expressed by the same genetic group of individuals in two different environments. Further, a correlation coefficient is estimated by obtaining the ratio between covariance of the two variables and the product of the standard deviations of the same variables. This is known as the product moment correlation coefficient. Similarly, the genetic correlation of the two traits could be the product moment correlation coefficient if the genetic covariance and genetic standard deviations of the traits could be measured directly (if possible) or indirectly and, therefore, could be written as Usually, the genetic covariance and variances are indirectly estimated by using the analysis of variance and covariance technique. This 31 technique is straightforward and simple when the traits are measured in the same individuals. Estimating the co-variance between traits when unequal frequencies of observed data in different individuals belonging to the same genetic group are considered is a very complex procedure. In such cases, in this experiment, the method suggested by Magee (1969) was followed in estimating correlation between the genetic value of a sire to produce Offsprings selected on the basis of growth rate performance when receiving either rations containing raw soybean or those containing cooked soybean. The covariance of sire effects in raw soybean and cooked soybean rearing environments were calculated by using the following formula: CW SR§C '3 [(ZIij.) (XI-13..) ' (751...”251' ...)/nj:] “j ' 1 where: SR = average of sire-progeny performance in raw soybean ration SC = average of sire-progeny performance in cooked soybean ration i = rearing environment i - ration with raw soybean i' = ration with cooked soybean j = sires n - number of sires :I The formula for calculating genetic correlation is: § _ _ - GRGC — Cov SRSC \/ V(SR) v (SC) where: Cov SRSC = covariance between the progeny means of the sires in raw soybean ration (SR) and in cooked soybean ration (SC) 32 V(SR) = the between sire component of variance in raw soybean ration rearing environment V(SC) = the between sire component of variance in cooked soybean ration rearing environment The progeny means were computed using one-way analysis of variance within rearing environment by line combination. The set program for hierarchical analysis written by Kratzer (1969), Antigenic Lab, Iowa State University was modified by Flangan (1969) for use by the M.S.U. Dairy Department in estimating heritability. The same program was further modified in the subroutine input for the present experiment to estimate the heritability and sire—progeny means within line by rearing environment. The between sire variance components within line by rearing environment were obtained from the same program through the 3600 Fortran computer, while estimating the heritability of the trait within line by rearing environment combination. Because of the negative estimates for the between sire variance components (Table 16) the intra-generation genetic correlation was not computed. The between sire variance com- ponents estimated within re3pective rearing environment as well as the covariance of the sires between rearing environments were pooled over all generations to estimate the genetic correlation between traits on the two rearing environments. C. Two-way Analysis of Variance Using Robertson's (1959) Fixed Model Concept: The expected mean squares for the biological relations with regard to the genetic correlation between and within genetic groups are presented in Table 3. Robertson (1959), working with a 33 Table 3. Expected mean square (Robertson, 1959). Sources df Expected M.S. Designation Rearing environment P - 1 Not relevant Between groups N - 1 l—t + nt [1 + (P-l) rg] A Group x environment (N—l)(p—l) l—t + nt (1-rg) B Remainder Np(n-l) l-t C A-B r G = A-C + (P-l)(B-C) A-B If P = 2’ rs = A+B - 2c 34 fixed model, used the standardized mean square values to compute the genetic correlations and computation as follows: A A - B re ’ A + B - (P-l) 20 when: A mean square due to genetic group B = mean square due to group by environment interaction C = mean square within genetic group P = number of environments, if P = 2 f = A - B G A + B - 2C Standard Error of the Genetic Correlation This is computed according to the formula suggested by Falconer (1960) and is a very approximate estimation. 2 1" c o(h2x) 0(h2y) 2 2 VFE" h x h y 0(rG) = Where: 0(h2x) = standard error of heritability for trait X o(h2y) = standard error of heritability for trait Y h2x = heritability estimate for X-trait h2y = heritability estimate for Y-trait Estimation of "Heritability: In a selection experiment the estimation of the pOpulation parameter, the heritability, is one of the main objectives. It may be 35 estimated by weighting the "true response to selection" over the "selection differential" or from the intraclass correlation coefficient computed from sires and dams component of variance for the selected trait. The former is usually known as the realized heritability. The "true response", in a selection experiment, is defined as the unit of progress that has been made in a given population for the selected trait from one generation to the next and the "selection differential" is the difference between average of the phenotype of the parents selected from a population and the average of the same population. This may be expressed as follows: where: M = difference between the mean of the offsprings and the mean of the population from which their parents were selected (P — P) - selection differential as defined in the text th = estimated realized heritability within generation In a bi-directional selection experiment, the ratio between cumulative response (upward and downward) and cumulative selection differential (up and down) is the best estimate for the realized heritability within a given generation. Similarly, over several generations when the cumulative response is regressed on the cumulative selection differential, the best estimate for the realized herita- bility is obtained for the given population. Thus, the realized heritability for the F and F4 generations 1 were calculated (Table 13) but for the F2 and F3 generations, it was not 36 possible to compute directly unless an estimated value was generated for individual sire or dam brought from the cooked soybean-ration- environment to become parents for subsequent generation. Heritability Estimate from Intra— class Correlation Coefficient The data for four groups of rearing-environments by line combina- tions (i.e., high-line raw-soybean ration; low-line raw-soybean ration; high-line cooked-soybean ration; low-line cooked-soybean ration) within a given generation were analyzed by one-way hierarchical analysis of variance. The program written by Kratzer (1969), Antigenic Lab, Iowa State University, for distribution, was obtained by the Department of Dairy Science, Michigan State University and further modified for CDC 3600 Fortran by Flangan (1969). The sub-routine input was written to analyze the above data within treatment combinations and pooled sex. The estimated parameter from different components of variance is expressed by the following notations: . 2_ 4S (1) hE-0+D+S (ii) hE2=_4_13___ 0+D+S . 2_2(D+S) (111) hE-0+D+S where: hE2 = heritability estimated for intraclass correlation coefficient S = between sires component of variance D = between dams component of variance 0 = between offspring (sires or dams) variance 37 The standard errors of these estimates from sires, dams and combined parents were calculated by the method of Dickerson (1969). The heritability estimates were then pooled within rearing environment over generations for obtaining the best estimate of the heritability for body size at 4 weeks of age in a given population under specific rearing condition. Expected Phenotypic Correlation Be- tween Selected and Correlated Traits The phenotypic correlation between two traits is expressed by the following equation: A I 2 ' 2 i I r = h h r _ 2 _ 2 plp2 l 2 Gle + (1 h) (1 hz) rE1132 where: h12h22 are the heritabilities of the respective traits and rG162 and rElE2 are the genetic and environmental correlation for the traits. When the environmental correlation is zero or does not exist (as in this experiment), the phenotypic correlation is a function only of heritability and genetic correlation of the traits. Therefore, the expected phenotypic correlation could be computed by multiplying the genetic correlation by the estimated heritability of the traits within the respective rearing environments (raw soybean and cooked soybean) for the given population. Figure 1 represents the biological paths involved in inheritance of the genetic material contributing to the similarities on phenotypes between parents and their offsprings. Correlation or variances are 39 used to describe the biometric relations between individuals. Lush (1948) described and discussed the use of the path diagram to estimate the various relations between relatives. In this specific situation the correlation between the genetic ability of a sire to produce offsprings for performance when receiving a raw soybean ration and his genetic ability to produce progeny for performance when receiving a cooked soybean ration has been illustrated by the possible paths (Figure 1). Though the sires and dams are the same in both rearing conditions their abilities to produce offsprings for specific environments have been separated and connected by double headed arrows indicating the genetic correlation between these two abilities for the respective phenotypes. On an average, over four generations, one sire was mated to 4.2 dams in the high line and 5.6 dams in the low line and average number of full sibs per dam was 7.6 in the high line and 10 in low line, respectively. The path between the genic value of a parent and its offspring is 1/2 in a random mating non-inbred population. The path between genic value of the offspring to its own phenotype is V[ha: The relationship with regard to the Figure 1 can be deduced as follows: 2n nh2 2 — ‘ 1” h 2 + (n-1)h2 " 4 + 2hT(n—1) 2 n . I a Equation 1. t 1/4 h 1 + (n-l)t where: t = 1/2 h2 and n = number of full sibs/dam 40 " 2 , 2 Equation 2: r =rG N t l ) + PO(R)P 0“) \/N(1 + (N-l) r') N<\/-t—'- \/ 1 )2] where N = number of POI N (1-(N-l)t') (I = 1 ..... N) = r-'(N+1) t' '] t L,1 + (N-l) t' ° G _ I-(N + 1) uh2 I. + 2h2 (n-l) . EC ‘ L? + 2hZIn-1) ' 4 + 2h2’(n-1) + (N—l) nhZ _ "(N + 1) nh2 E ‘ 4 + 2h2 (n-l) + (N—l) nhZ’ G (N + 1) nh2 £6 ‘4 + h2 [(N + 1) n-2] MATERIALS AND METHODS From the review of literature, it could be concluded that the raw soybean in a diet when fed to chickens depresses the body size but with a mechanism not solved unequivocally. Regardless of species, body growth is a trait sensitive to external and internal environmental in- fluences. If chickens are reared in stressed environments (e.g., nutritional or physical), body growth is depressed. Nevertheless, attempts to establish genetic variability of body size by selecting individuals from stressed nutritional environment have been reported by many workers (e.g., diet deficient in methionine [Hess $3.31., 1962; Lepore, 1965], deficient in lysine [Enos and Mbreng, 1965; Godfrey, 1968] and containing depressant factors, thiouracil [Marks and Lapore, 1968a, 1968b]) with controversial conclusions. However, the growth rate of an individual is moderate to highly inherited and, therefore, under Optimum environmental conditions the body size by weight could be increased or decreased (Siegel, 1962; Nordskog and Festing, 1962) by mass selection. When the environment is different from the optimum (either man-made or natural) the growth or the body size still could be increased or decreased depending upon how heritable this "body size" is in the given environment. The concept of regarding one trait as two when it is measured in two environments has been introduced by Falconer (1960) because of the underlying complex physiological mechanisms 41 42 reacting differently in different environments. Therefore, the body size measured in different environments should be regarded as different traits. The depressing factors present in the raw soybean to cause a "depressed growth" and the inherited potential for "growth" of an in- dividual are two confounded traits (in absence of the knowledge re- garding the metabolic functions of these depressant factors), if the "body size" is the chosen parameter to measure the effect of the raw soybean. Therefore, an increase in body size by selection of the individual's own performance under a dietary environment with raw soy- bean, would be regarded as a "growth-tolerance" pe£_§e_which may be defined as a net growth resulting from the depressant effect due to raw soybean in a diet and the genetic improvement for body size due to selection. Similarly, a rapid decrease in body size in the same environment may be called a "growth-intolerance" pe£_§e_which may be defined as a total depressed growth because of simultaneous negative forces, the one from the negative selection for body size and the other from dietary depression. In other words, a divergence of body growth by weight, due to selection in a diet with raw soybean, may be considered as tolerance or intolerance to the raw soybean. So, based upon this hypothesis, a two-way selection program for high and low body size in a raw soybean dietary environment was initiated in 1965 using the chicken as a research animal. Since the growth depressing factors present in raw soybean are heat labile, a group of close rela- tive (full-sibs or half-sibs) when permitted to express their body size in two similar dietary regimes, one with raw soybean and the other having cooked or treated soybean, would be expected to perform 43 differently. The difference in the average body size of these two groups of relatives coming from the same sire and dams would measure the variability due to genotype by environment interaction. Therefore, this estimated interaction parameter could be used to predict the cor- related response of the genotypic performance, which would be carried over to an environment other than the one where it was primarily measured. To follow up the selection program, the following assumptions were made. 1. The additive genetic variance for body growth of birds fed diets with raw soybean should be approximately 41 percent. (The median heritability for body growth of birds fed a "normal" diet has been reported at this level.) 2. Mass selection would be effective if the above assumption is correct. 3. The coefficient of variation for body weight of birds fed diets with raw soybean and those fed diets with treated soybean should be the same. 4. The palatability of the diet containing raw soybean as compared to that of similar diets with treated soybean should not cause a differential feed intake. 5. After absorption from the gut, the utilizable protein and the depressant factors present in raw soybean should in- fluence the body growth independently (assumption of linearity for mutually exclusive events). 44 The genetic correlation of the two dietary environments should be nearly 1 if the body growth is affected by the same set of genes and low if it is not affected by the same set of genes (assumption of pleiotrOpy). GENERAL EXPERIMENTAL PROCEDURE A. Base Population In 1964-65, from a project involving a study of pancreatic hypertrophy in a group of Cobb strain white Plymouth Rock male chicks which had received a ration containing raw soybean during the accelerated phase of growth (0-4 weeks), 18 males were saved. During the same year, 200 one-day-old pullet chicks of the same strain were received from the breeder. The feed program for the pullet chicks was the same as that of the males. The base population was assumed to have no in-' breeding and to be heterogenous with regard to the trait (body weight) chosen for selection. B. Nutritional Rearing Environment In Table 4 is shown the two rations used in this experiment for chicks from one day until 4 weeks for all the generations. The two rations were iso-nitrogenous (21% protein) and isocaloric (=2977 Kcal/ Kg). The ingredients, other than corn and soybean, in both rations were exactly the same in quality and quantity. The soybean flakes in Ration #2 had the "growth depressant factors" whereas the dehulled soy- bean meal in Ration #1, due to adequate heat treatment, did not have them. The soybean flakes were contributed by Central Soy Company, Decatur, Indiana, and had been fat extracted at 69° C to permit flaking and were known to retard growth slightly less than does the raw soybean. 45 46 Table 4. Experimental rations used for the chicks from one-day—old until four weeks of age. Ingredient Ration #1 Ration #2 (7.) (7°) Ground corn (8.5% protein) 61.21 64.10 Defatted soybean f1akes* ----- 24.55 Dehulled soybean meal (50% protein) 27.44 ——-—- Alfalfa meal, 20% protein 2.00 2.00 Vitaproil, 57% protein 3.00 3.00 Dried corn distillers solubles 2.00 2.00 Dried whey 2.00 2.00 Salt 0.50 0.50 Ground limestone 0.90 0.90 Dicalcium phosphate 0.70 0.70 Nopcosol M-5** 0.25 0.25 Amprol*** x x Total 100.00 100.00 *Soybean flakes contributed by Central Soya Company, Decatur, Indiana. Soybeans are tempered by steam to permit flaking and removal of hulls. They are fat extracted at 69° C. The flakes retard growth slightly less than raw soybeans. **Nopscosol M-5 contributed by Nopco Chemical Company. ***Coccidiostat. 47 Since the commercial soybean meal was devoid of bulls and fats, it was imperative that the raw soybean should be dehulled and fat extracted in order to be a comparable ingredient in the Ration #2 with soybean meal in Ration #1. These growth depressant factors are heat labile (248° F for 15-20 minutes or autoclaving at 15 pounds of pressure for 30-45 minutes); however the heat treatment of the soybean flakes at 69° C (124.2° F) could be assumed not to inactivate all the growth depressant factors present in raw soybean flakes and should, therefore, represent the "raw soybean" for all practical purposes in view of the selection experiment. The rations with raw (soybean flakes) and cooked (soybean meal) soybean were fed to alternate hatches of chicks; i.e., the first batch received raw soybean, the second hatch received cooked soybean, etc. A minimum of two hatches was obtained from each group of sires and dams which had been reared and selected in an environment where they received raw soybean in their diets. Chicks received one of the two rations from one day until 4 weeks with the same brooding procedures regardless of the ration they were on. After four weeks, all the selected chicks of either sex, irrespective of previous ration received, were given in sequence the following "regular" rations for— mulated in the Department of Poultry Science, Michigan State University: 1. Chick grower - until 8 weeks 2. Pullet developer - until 20 weeks 3. Layer—Breeder '63 - 20 weeks and after Usually the pullets and cockerels were individually caged at or before 20 weeks of age, prior to starting of the specific mating programs by artificial insemination. 48 C. Selection and Mating The bi-directional selection for body size (high and low) of chicks at four weeks of age in the raw soybean ration fed line was based on the individual's own phenotypic performance. In principle, no chicks were selected from the cooked soybean ration fed line on their own phenotype though some males and females from F1 and F2 offsprings were permitted to become parents for subsequent generations on the basis of their full-brothers' or full-sisters' performance on the raw soybean ration. From the base population five males and 24 females for each high and low body size at four weeks of age were selected and designated "high-line" and "low-line", respectively. The other chicks were dis- carded. Before the mating program started, one male and one female from the low line died and could not be replaced. Therefore, the F1 offspring generation was produced from five males and 24 females in the high line and 4 males and 23 females in the low line. The decision to make the selection on body size at four weeks was made when the males of the base population were already about 8 to 10 weeks of age. These males had been weighed when three and one-half weeks of age. So, the individual records for the 18 males were pro- jected to four weeks by the French Curve, as suggested by Gill (1969) in calculating the averages for the base and selected population. Since the weekly records for body size were available for those males, the four-week body weight could be predicted with little bias. From the F1 offsprings, six males and 28 females for the high and low line, respectively were selected. One male and one female in 49 the low line did not reproduce and therefore, the F offsprings were 2 generated from six males and 28 females for the high line and five males and 27 females for the low line. Out of these selected males and females, one male and 13 females in the high line and 12 females in the low line were reared throughout on the cooked soybean ration (Ration #1). From the F2 offsprings, six males and 26 females for the high line and six males and 39 females for the low line produced the F3 generation offsprings. Among these males and females, two males and eight females in the high line and 13 females in the low line were reared on the cooked soybean ration. The selection of parents from the F offspring population was 3 made on the basis of mass selection from raw soybean as well as from cooked soybean ration rearing environments. Therefore, five males and 14 females for the high line and five males and 23 females for the low line were selected from the raw soybean ration fed line. Similarly, six males and 18 females and five males and 20 females from the cooked soybean ration rearing environment were selected for the high and low line, respectively to produce the F4 generation. The above procedure was used to estimate the effect of direct and indirect selection of parents on their offsprings' performance with reference to the parental rearing environments. This is discussed in the chapter entitled "Application". The mating system, in general, was purely on a random basis within line and care was taken not to permit any full-sib or half-sib mating in all three generations, subsequent to base population. On an J I" m ‘As .1. 50 average, 4.1 dams per sire in the high line and 5.6 dams in the low line on an overall generation basis were permitted to produce the offspring by artificial insemination (0.1 m1 semen per hen twice weekly). Care was taken to inseminate the hen twice weekly with semen from the same sire to insure adequate fertility and accuracy of pedigree record. In case the hen did not "pop" on the date of insemination, another attempt was made to inseminate her later in the week. Because of a limited population and matings restricted within line, it was necessary to calculate the increase in inbreeding coef- ficient for this population over four generations. On an average, 5.5 sires and 23 dams in the high line and five sires and 28 dams in the low line were the breeding individuals. Therefore, the coefficient of inbreeding increase per generation was .027 and .025 in the high and low line, respectively, from the following formula by Wright (1940): number of males AF = 1/8m + l/8f where: m f number of females The inbreeding depression was then calculated within line. The depression per one percent increase in inbreeding for the high line was '-.0206 i .0238 gms, evidently not significant; whereas, for the low liJie, it was -.1358 i .0086 gms, apparently significant. However, in thetlow line, 2.5% increase of inbreeding would likely depress the IHDdy size about .37 grams per generation and therefore, may not bias the Conclusion in the downward selection. .Q;__1ncubating and Brooding Within a given generation, when the flock reached at least 50 Percent production and continued to lay eggs averaging 57 gms or more, 51 the planned mating decision was carried out by artificially inseminating each hen with semen from a specific sire. Trapnesting the eggs of the individual hen was facilitated due to individual caging. The eggs were refrigerated at 60° F from time of laying to beginning of incuba- tion. At no time were eggs kept more than 15 days prior to incubation. The hatch effect was considered negligible due to consecutive sequence of hatching of two batches of eggs within a given month of the year. Further, because of the experimental design, any difference in body size due to hatching was confounded with the difference due to rearing environments. All eggs were incubated in Jamesway 252 incubators and at the 19th day of incubation, they were transferred to the hatcher until 22 days. Normal incubating procedures for pedigree hatching were followed. As soon as the chicks were dried, all were wing banded, weighed in grams and then intermingled between lines for brooding. Records re- garding fertility, hatchability and any abnormalities in hatching according to sire-family were maintained for all the generations. The chicks in the F1 generation were brooded in Petersime starting batteries and were weighed at weekly intervals. Unfortunately, some chicks in the F1 generation exhibited a leg weakness; consequently, succeeding generations were brooded on the floor. The weekly weighing also was suspended in succeeding generations to avoid extra handling of the chicks. All chicks were given water and feed ad_1ibitum during the brooding period. The brooding mortality was less than 10 percent for each of the first three generations and about 17 percent for the fourth generation. RESULTS AND DISCUSSION The average rate of gain per day, by weekly age interval, from hatching to four weeks of age for offspring from the parental genera- TJ V fir . tion of birds which had been selected for 4—week body weight when fed g with the ration containing raw soybeans is shown in Table 5(a). The pattern of mean body weight gain for four groups of chicks at weekly n. _ intervals until 4 weeks is shown in Figure 2. The divergence, towards 4-weeks of age, of the growth curve for lines receiving the raw soybean ration and a parallel curve for lines receiving the cooked soybean ration may indicate a differential rate of gain of birds receiving the two rations. From Table 5(b), ANOVA for rate of gain, a significant difference (P < .05) is indicated between lines of chicks reared on rations containing raw soybean whereas there is no significant dif- ference between lines of chicks reared on the cooked soybean ration. Regardless of rearing conditions the linear and quadratic increases in body size for all groups of chicks are highly significant (P < .01). In Table 6, the coefficients of variation for different groups of chicks (line x rearing conditions) are presented. As the estimates are below 20 percent in all cases, the scale effect between mean and ‘variance may be eliminated and therefore the transformation of data in this experiment is not necessary (Falconer, 1960) for the analysis of the results . 52 53 Table 5(a). Rate of gain in body size per day by week for different line x rearing environments. O-lst l-2nd 2-3rd 3—4th week week week week Total Highline — raw soy 4.9 11.9 17.4 19.3 53.5 Highline - cooked soy 6.0 15.6 21.9 23.6 67.1 Lowline — raw soy 4.1 10.0 14.6 17.7 46.4 Lowline - cooked soy 5.7 14.4 20.2 23.0 63.3 Total 20.7 51.9 74.1 83.6 230.3 Table 5(b). ANOVA. Sources df SS MS "F" Rearing env. x line 3 66.2 22.1 - C1 = HR vs. LR 1 6.3 6.3 5.65* C2 = HC vs. LC 1 1.8 1.8 1.6 C3 = Raw vs. Cooked 1 58.1 58.1 52.8** Weeks (time) 3 585.6 195.2 - SSL 1 556.0 556.0 505.4** SSQ 1 29.4 29.4 26.7** SSC 1 0.6 0.6 <1 ERROR 9 10.3 1.1 *P < .05 **P < .01 Body weight (gms) 54 550 High-cooked soy 500 Low—cooked soy 450 High-raw soy 400 Low-raw soy 350 300 250 200 150 100 ./ l 2 3 4 Weeks Figure 2.-—Growth curves from hatching to four weeks of age for offspring of the high and low line from the first generation to fourth generation receiving growing rations containing raw or cooked soybean. 55 .6 e. . 3.... .ru . .hl. a»: NH. mm eom NH. Nm one mH. ee NmN mH. oe eoe eo NH. oN mHe NH. Ne ewe NH. em eNN NH. oN Nam mu NH. oe OHm NH. Ne mew NH. em NHN eH. on see No HH. Nm ewe NH. Ne NHm mH. we con NH. em ONe Hu .>.o .o.m sees .>.o .n.m ewe: .>.o .o.m sees .>.o .n.m new: eexooo 30H eexooo eme are 30H rem emHm mp AmEmv muswfimB mvon xmo3tusom pow coauMNum> mo .mcofiumuonow was nomad ucofiofimmooo new cowumfi>ov cumuaMum .amoz .0 mHQMH 56 Genotype x Environmental Interaction Table 7 shows the regression coefficients with standard errors for various variables included in the model as discussed in the text. The means for different lines by rearing conditions adjusted for line, rearing environment, and their interaction effects by sexes and pooled sexes are presented in Table 8. The ANOVA for these means are shown in Table 9a, b, c and d for generations one to four, respectively. The means for the four groups (line x rearing environments) have been plotted in Figure 3 to illustrate the changes from one generation to the next. Between lines, between rearing conditions and line x rearing environment (genotype—environment) interaction are highly significant (P < .01) in all generations. are graphically presented within The achievement of a high assumed to be due to the average types segregated in the high and conditions (having or not having soybean). So the four groups of two rearing conditions) could be ‘ The genotype-environment interactions generations in Figure 4. or a low body size by selection was effect of "better" or "poor" geno- low line respectively in two rearing growth depressant factors in raw treatment combinations (two lines x arranged as follows: ‘ Highline Lowline (better genotype) (poor genotype) Ration with cooked soybean HC LC (good dietary environment) (1) (3) Ration with raw soybean HR LR (poor dietary environment) (2) (4) ‘ Class 1 2 3 4 'Trt. group HC > HR > LC > LR 57 Ho. y mas HI u onEwm H+ u mHm: NI n mcHH Sou H + u eeHH emHm HI u A.>amv GOHumH amonhom pmxooo H+ u A.>amv GOHumu ammnhom 3mm "mHmhamcm moumsvm unwed mnu aH wow: mmpoo enema H SN- eeeNN H 2.? eeNmN H HN.HH- eerN H Re. New x cam «fies H 8.2 eeeNN H eoe reNeN H wmaH eroeN H Nee and a. cam «reme H 8.8 ENNN H Ntee ENHN H NNNN .336 H No.2 83 .326 H eeeH ereNN H He.eH erNeN H eNeH eemmN H NNHN New .236 H Rem- reeNN H 3.3. eeNeN H Sew- TEN H 3.8- $5 ee.ewm mm.msm Ne.mme om.ome .emeoe HHeee>o .u GOHHNHMGUU m COHHQHGGGO N COHHMHUGOU H GOHUMHUGOU .uano3 hvon xomstuaom you name onu ou muoommo mcHustHucoo moHanum> mDOHum> How .m.m :uHs mucmfiowwmmoo =0Hmmmuwmm .N mHan 58 Table 8. Mean body weights (gms) adjusted for line, sex, rearing, environment and their interaction effects at 4 weeks of age by generation for the high and low line chicks receiving raw or cooked soybean in their ration. Generations Po 1 2 3 4 Direct Selected Trait (body weight on raw soybean ration) High Raw Male -- 432 400 395 412 Female -- 402 394 373 400 Combined sex 403 417 397 384 406 Low Raw Male -- 377 316 294 264 Female -- 346 310 272 252 Combined sex 403 361 313 283 258 Correlated Trait (body weight on cooked soybean ration) High Cooked Male -- 549 563 517 513 Female -- 493 512 473 465 Combined sex -- 521 537 495 489 Low Cooked Male -- 528 534 440 418 Female —- 472 483 396 370 Combined sex -- 500 509 418 394 ¥__ 59 Table 9. ANOVA of four-week body weights (between lines and between environments) by generation. (a) Generation 1 Sources df SS MS "F" Rearing environment 1 l4l37l3.9l l4l3713.91 545.47*** Sex 1 183933.22 183933.22 70.97*** Line 1 61322.19 61322.19 2.85 Sires/line 7 150368.88 21481.27 5.42** Sires/high 4 50282.39 12570.59 -- Sires/low 3 100086.49 33362.16 —- Dams/sires/line 38 150576.59 3962.54 l.53** Dams/high 19 54188.38 2852.02 -- Dams/low l9 96388.21 5073.06 -- Rearing environment x line 1 29833.16 29833.16 11.51** Rearing environment x sex 1 16621.40 16621.40 6.41** Line x sex 1 2378.38 2378.38 <1 Rearing environment x line x sex 1 11.89 11.89 <1 Error 417 1080740.77 2591.70 Total 469 3279744.25 **P < .01 ***P << .01 Table 9. 60 environments) by generation. (b) Generation 2 ANOVA of four-week body weights (between lines and between Sources df SS MS "F" Rearing environment 1 3445077.13 3445077.13 Sex 1 98674.58 98674.58 34.65*** Line 1 235073.93 235073.93 24.32** Sires/line 9 86986.08 9665.12 1.32 Sires/high 5 48171.66 9634.33 -- Sires/low 4 38814.42 9703.61 -— Dams/sires/line 44 322614.74 7332.15 2.57** Dams/high 22 177106.17 8050.28 -- Dams/low 22 145508.57 6614.03 -- Rearing environment x line 1 94977.94 94977.94 33.35*** Rearing environment x sex 1 63440.61 63440.61 22.29*** Line x sex 1 1035.35 1035.35 <1 Rearing environment x line x sex 1 2141.92 2141.92 <1 Error 528 1520610.23 2879.9 Total 588 7472867.76 **P < .01 **P <<.01 61 Table 9. ANOVA of four-week body weights (between lines and between environments) by generation. (c) Generation 3 Sources df SS MS "F" Rearing environment 1 2087521.19 2087521.19 741.9l** Sex 1 148217.44 148217.44 52.68** Line 1 406681.55 506681.55 17.12** Sires/line 10 238284.87 23828.49 3.02** Sires/high 5 123325.18 24665.03 -- Sires/low 5 115959.60 23191.92 -- Dams/sires/line 53 417660.29 7880.38 2.80** Dams/high 20 218051.69 10902.58 —- Dams/low 33 199608.60 6048.74 -- Rearing environment x line 1 20072.37 20072.37 7.13** Rearing environment x sex 1 16609.30 16609.30 5.90** Line x sex 1 1412.76 1412.76 <1 Rearing environment x line x sex 1 13156.32 13156.32 4.66* Error 598 1682596.30 2813.71 Total 668 6448199.85 *P < .05 **P < .01 62 Table 9. ANOVA of four-week body weights (between lines and between environments) by generation. (d) Generation 4 Sources df SS MS "F" Rearing environment 1 575744.18 575744.18 249.09** Sex 1 46637.39 46637.39 20.18** Line 1 350624 350624.33 83.03** Sires/line 8 33781.13 4222.64 1.23 Sires/high 4 20348.64 5087.16 -- Sires/low 4 13432.49 3358.12 —- Dams/sires/line 27 92482.85 3425.29 l.48** Dams/high 9 26109.30 2901.03 -- Dams/low l8 66373.55 3687.42 -- Rearing environment x line 1 33771.87 33771.87 14.6l** Rearing environment x sex 1 16181.39 16181.39 7.00** Line x sex 1 929.61 929.01 <1 Rearing environment x line x sex 1 4364.76 4364.76 1.89 Error 229 529293.51 2311.32 Total 271 2652872.06 **P < .01 at four w b k C 4_ Mean bod geks of age y weight (gms) 560' N 0 co 0 c> c? 360‘ 320‘ 280' 240 63 Raw Soybean (Direct) Cooked Soybean (Correlated) -— — — — -- —— HC = High Line Cooked Soybean ‘,\\ LC = Low Line Cooked Soybean ’ H ’ \ HR = High Line Raw Soybean ev" ‘~‘\ LR = Low Line Raw Soybean a—“". \ “’v” \ \ \\ _____ —'"‘"'HC \ \ \\ \ \ \. \ \ \ \ 406.8 - 2.7x ‘«s HR LC = 2.7 + 4.1 HR '— = ** bLR 36.8 j; 2.8 LR i '2 ’3 ’4 T Gene rations Figure 3.--Mean response for selected and correlated traits (four- week body weights) for the high and low line offspring fed diets con- taining raw or cooked soybean. 64 150 Raw Soybea Diet [.1 N S’ --— —'— — — -- -' - Cooked Soy ean Diet 88 \o 5’ U) C 4 (gms) between the high and low lin ox F’ l l I I let Generation 2nd Generation Absolute difference of 4—week body weight H O h 1. H L 150I g.» 0 N 9! 9’ Ox 9’ Absolute difference of 4-week body weight (gms) between the high and low lines u: c> 3rd Generation 10 H L Figure 4.--The effect of genotype—environment interaction as a devia- tion from overall constant by line and environment within a generation basis. 65 The expected ranks for the performances of the genotypes in two rearing dietary environments are listed in the parenthesis within re- spective cells. A statistical significance in the shifting of these ranks would, therefore, indicate interaction, though it is not a pre- requisite. A differential degree in response of the genotypes to dif— ferent environments would make such interaction significant even though the rankings remain the same. With reference to Haldane's (1918) six possible relations of different phenotypes (given by two homozygous genotypes for a pair of genes and two environments) as listed by Mather and Jones (1958), it is noted that where difference between genotypes in both environments is small as compared with difference between environments, the relation of the phenotypes would be in order of the ranking as l > 3 > 2 > 4 (assuming here that two lines are homozygous genotypes for one pair of .genes). Similarly the order of ranking would change when the environ- mental difference is smaller than the genotype differences and in that case the phenotypic ranks would be 1 > 2 > 3 > 4. Therefore, at least in the two above specific situations of rankings, because of the presence of the genotype-environmental inter- action, the breeder is restricted in predicting the responses of the genotypes with respect to environments. However, the breeder is still interested in knowing the genotype-environment interaction in order to be able to formulate breeding plans with regard to strains and environ- ments (location, diet, etc.) and to select the very efficient strain (genotype) adaptable to the specific environment, if possible. The adjusted means for the four groups of treatment combinations (HC, HR, LC, LR) in this experiment, when arranged on a within 66 generation basis, give an order of rankings of l > 3 > 2 > 4 from the first to the third generation; whereas the order in the fourth genera- tion is l > 2 > 3 > 4 (Table 8). Referring to Table 7, in generation 1, the between-environmental differences (2 x regression coefficient for environmental effect) is 121.5 as compared to 38.0 for the between line differences (2 x regres— sion coefficient for line effects). Due to a linear decrease in environ- mental effect starting with the second generation, the difference be— tween environments is 109.6 at the fourth generation; and, similarly, because of increase trend of line effect, the differences between lines over environments is 121.6 at the fourth generation. Therefore, the ranking as discussed was expected to be in the order of 1 > 3 > 2 > 4 for the first three generations and l > 2 > 3 > 4 for the fourth genera- tion depending upon the magnitudes of the differences of line and environmental effects as pointed out. Since the line or environmental effects (regression coefficients) become variables with regard to generation, the regression of line or environmental effects on generations would estimate the true changes of these effects in estimating the average body size per generation in the high and low line chicks fed with rations containing raw soybean. Thus, the A line and A environmental effects when calculated over four and three generations, respectively, (by regressing the least squares estimates on generation) are 14.2 and —14.7. Therefore, the net effect on high line chicks due to these two independent effects would be -.5 per generation whereas for the low line chicks, the net effect is -28.9 per generation. 67 This situation may be illustrated by assuming the line and environmental effects by single arrows (0A, OB, EE') directed as shown in Figure 5 in both lines. Thus, the resultant arrows (0C and OD) (net-effects) may be due to the balancing effect of the line and the environment in the high line and a cumulative effect of the line and environment in the low line. This could be a plausible explanation to account for the upward slow progress and downward rapid progress of the mean per generation for the high and low line chicks, respectively, when fed the ration containing raw soybean. However, when the adjusted means (true responses) for the high and low line chicks fed with rations containing raw soybean are directly regressed on generation, the estimated changes of the means per genera— tion for the two lines are -4.6 :_6.9 and -33.9 i 3.7, respectively (Figure 3), with a divergence of 29.3 gms per generation (Figure 6). These two estimates are comparable to -.5 and -28.9 estimated by inde- pendent regression of line and environment effect on generation, since in both estimating methods the divergence betwen high and low line per generation is nearly 29 (Figure 6). Since the regression of true means (adjusted) on generation for the high line is —4.6 iL6-9’ this indicates that the straight line is not a fit one to account for the change in mean per generation because of the magnitude of the standard error for the regression coefficient. A polynomial response curve (parabolic function of the change of means with respect to time) might verify the A mean per generation. Thus, a quadratic curve was fitted to the four points observed in the four generations and the following was the estimate obtained (Figure 7): y = 465 - 57.1 x +1o.5 x2 68 b(LSE-G)H = 0A = Upward selection effect. b(LSE-G)L = OB = Downward selection effect. b(LSE-G)E = EE' = Environmental effect. E CC = Resultant effect in upward selection. 0D 8 Resultant effect in downward A selection. ;, LSE - Least squares estimate for high f‘ (H) and low (L) and environment \rf" “ (E). \v \ 0 I x \. ‘\‘§§ \ \e \\ f e \I “3° \ .0 \ C Figure 5.—-Independent effect of genetic and environmental forces in determining the net change in four—week body weight per generation for than lie-IA]... A...) 1..-- 11...; “Canv-Inn vadn4iv4nh 9 (14¢? nnnf'n‘fn‘fna ram anxrkann ' 69 0 00 m “3 160' CD .5: CD a) I3 ‘7 140 u to >x at: on 000) 8'3. @0120 0) Se Hm r-IH 360 25100- t: '64-! an: mu {.1 .Co 000 ~I-I ‘8 80 «DH ,cu um M NH Oct) HS (0 .- 50'“ 60 u VH 03 0"!) an) 314-: a, 40- M M H 'U u f. H 20~ Q) 3 >x 'U 0 m o - . - ' 1 2 3 4 Generations Figure 6.--The effect of selection over four generations as divergence between the high and low line offspring receiving ration containing raw soybean. 7O 430II 41d 0 \ \, J? 40 \\\ 'F 465 . ‘ 5 \s ‘ ?'-ZX 3L , ’/ ”e“ gm 37 U .c an ’3 K i \ v8 35 ~ .0 s x \ a) ' "I a) J» '3 33 0‘ ‘5 ~5ch ‘3 \ Ln \\ \QT ‘59 3101) \+ \ \. \ \ \\ 290$ \ \ II V 270 \ \ \ - \ \ O 250 ‘ ‘ ‘ ‘ l 2 3 4 Generations Figure 7.-—The fitted quadratic and linear lines for the high and low line mean 4-week body weights (gms) of offspring fed raw soybean ration over four generations. 71 Thus, we see that fitting a straight line accounts for only 18 percent of the variability in y (19§;§) and fitting a quadratic curve 586 441.0 . accounts for 93 percent [ 586 ] + 105.8 . Since the straight line function in the low line accounted for 97 percent of the variability in y, the possibility for any curve other than a linear one is out of the question. Therefore, it may be concluded that a linear change in the low line and a quadratic change in the high line for the means with regard to generation (time) could be the acceptable hypothesis within these four generations. Asymmetry of Response Referring to Table 10, the cumulative response for the high line is far below the expected response and the reverse is the situation for the low line where the observed response was more than was expected (Figure 8). The environmental effect is cumulated in the low line on the genetic effect for low body size and this effect is, probably, counteracted by the genetic force due to selection for a high body size. Because of a linear change in the low line and a quadratic change in the high line, an asymmetric response was found in this experiment. Falconer (1960) discussed five causes of asymmetry of response in two- ‘way selection experiments. Since "genetic asymmetry" due to directional «dominance or directional gene frequency and selection heterozygotes are not the causes for asymmetry in the first few generations, the (ather important cause, the selection differential, may be the one which ¢ieserves discussion. Unequal selection differential for the high and i10W'1ines are contributed by three factors, i.e., (l) differential Iiatural selection with artificial selection, (2) differential fertility 72 0.0NI 0.0eH 0.0NH 0m.0 0e NN.H 0e.H 0N.0 0H.N 00.0 e 0.0NI 0.0NH 0.e0 0m.0 em 0e.H N0.H Nm.0 00.H 00.0 m 0.eNI 0.00 0.00 0m.0 N0 we.H eo.H 0m.0 Nm.H No.0 N 0.0I 0.Ne 0.0m 0m.0 m0 em.H NN.H HH.0 0m.N NN.0 H .aoo mmIHIHIHmm 0.0HH 0.m 0.NNH 0m.0 00 00.H me.H NH.0 00.H 00.0 e 0.00H 0.0HI 0.0m 0m.0 0N em.H 00.0 00.0 ww.H 00.0 m 0.00 0.0I 0.Nm 0m.0 mm 00.H He.0 mm.0 0N.H 0H.0 N 0.mH 0.eH 0.0N 0m.0 0m me.H m0.H NH.0 0H.H 0N.0 H .awo meIHHIHmHm mocmummmHm o>HumHsaso m>HumH=a=o m: m 0\n 0\N mHmEmm 0\N onz 0m>uom00 wouomnxm N b .o>< uncommom n\u 0am wo>mm coHuomum .mGOHumuoamw 0cm wmaHH an muanms >000 xooslusom pom uncommon 0>Humanabo 0m>ummno 0cm wouoqum .0H manma 73 Expected response --- — — - - - - - Observed response 560" J ' High line 520 480 I ’8 E 8" 4..) €3440" ‘H o 3 i? ,.—"”'”’”~"‘~ g flu. \ \~ "High line .14 400 r \' ~ ” x I 3 \\ ‘T‘ ’ ’ .3 \ I" a \ FL. «I 360 \\ \ \ Is \ \\ 320 \ S.‘ \\ \ \- 280' \ ‘ \ \ Low line “- \\ °Low line 240 If ‘ ‘ i 1 2 3 4 Generations Figure 8.--Observed and expected response to direct selection for four—week body weight in the high and low line offspring receiving the raw soybean ration. 74 and (3) scale effect. In Table 13 the observed and expected selection differentials for first and fourth generations are presented. The observed and expected selection differential in Po are quite agreeable with one another in both lines whereas in 63 the observed selection differentials are about 60% and 80% in the high and low line, respec- tively, of the expected selection differentials. Differential percent fertility was present between the lines (Figure 10) with an increased trend over generations for both lines. Since the coefficient of variation was below 20% in either rearing environments, the scale effect "1' Edi... LIZ-rid. I may be ruled out. Inbreeding depression and maternal effect are two other possible causes for asymmetry of response. Non-existence of inbreeding depres- sion in the high line and very negligible depression in the low line (.37 gm per generation) may not be very important in accounting for the observed asymmetry. The maternal effect in the high line is 9 percent as against 3 percent in the low line contributing to the phenotypic variation. Falconer (1960) pointed out that the maternal effect is not a valid point to support the asymmetry since, by considering this effect, only the shifting of one character to another is favored and it does not solve the specific problem of determining the cause of asymmetry. ngulation Parameter Heritability, Realized Heritability and Correlated Heritability A summary of the heritability estimates for body weight at four weeks of age on pooled sex by line and rearing environments on a within generation from between-sire, between-dam and combined parents compo— nents of variance is shown in Table 11. The estimates for the high 75 Table 11. Heritability estimates for four-week body weight. Direct Selected Trait (Raw Soybean body weight) Generations l 2 3 4 High Raw 43 “NS 0.23 i 0.29 0.27 i 0.37 0.14 i 0.39 .42 i 0.57 4D -07+030 051+040 132+058 O6+O63 F+D+S . _ . . _ . . _ . . _ . 2(D+S) ————F+D+S 0.08 i 0.18 0.39 i 0.23 0.73 i 0.29 .18 i 0.32 Low Raw 48 “NS 0.99 i 0.93 0.21 i 0.29 0.54 i 0.43 .04 i 0.27 4D 0 25 + 0 24 0 48 + 0 30 0 32 + 0 22 22 + 0 46 F+D+S . __ . . __ . . _ . . _ . ADJ-LS)— 062+O47 034+019 043+023 13+021 F+D+S . _ . . _ . . _ . . _ . Pooled Within Generations 0.61 i 0.48 0.24 i 0.23 0.34 i 0.29 23 + 0.31 0.09 .t 0.19 0.49 i 0.25 0.82 i 0.30 08 i 0.39 0.35 i 0.25 0.36 i 0.15 0.58 .t 0.18 16 it 0 16 Pooled Within Rearing Environment 0.35 i 0.17 0.37 i 0.14 0.36 i 0.09 76 Table ll--Continued Correlated Trait (Cooked Soybean body weight) Generations l 2 3 4 High Cooked :fims 0.43 i 0.55 0.41 i 0.55 0.48 i 0.53 0.36 i 0.70 22mg 0.38 i 0.48 0.81 i 0.52 0.69 i 0.41 -.03 i 0.71 F511;? 0.41 i 0.31 0.61 i 0.32 0.58 i 0.30 0.17 i 0.32 Low Cooked £1378 0.23 :t 0.38 0.04 i 0.16 0.39 i 0.39 -.07 i 0.33 :Ems 0.32 i 0.34 0.31 i 0.27 0.49 i 0.31 1.01 i 0.63 fig? 0.27 i 0.22 0.18 _+_ 0.13 0.44 i 0.23 0.47 i 0.26 Pooled Within Generations 0.33 i 0.33 0.22 i 0.28 0.43 i 0.33 0.14 i 0.38 0.35 i 0.29 0.56 i 0.29 0.59 i 0.25 0.48 i 0.47 0.34 i 0.19 0.39 i 0.19 0.51 i 0.19 0.32 i 0.20 Pooled Within Rearing Environment 0.16 O N 00 l+ 0.49 i 0.17 0.39 _t 0.09 77 line in the rearing environment where they received raw soybean in the ration range over four generations from 0.14 to 0.23 due to the between- sire variance component; whereas, in the low line they vary from 0.04 to 0.99. Similarly, from between-dam component of variance, these estimates in the high and low line range from -.O6 to 0.32 and from 0.22 to 0.48, respectively. These estimates are subjected to a sampling error twice that estimated from the pooled parents component of variance. Therefore, the best estimate of heritability on a within generation for the four—week body weight regardless of lines or rearing environments may be considered as twice the fraction of sire and dam variance, al- though such an estimate is inflated by various unknown amounts of non- additive, sex-linked and maternal effects (Kempthrone, 1957; Falconer, 1960). The ANOVA for the means of the estimated heritability due to full-sib correlation (pooled h2 computed from sire and dam component of variance) for line-rearing condition by generation is presented in Table 12 and plotted in Figure 9. Neither the line by rearing condi- tions nor the generations are significant statistically. The quadratic (ieclining trend of the h2 over generations is not significant. This Vvas probably due to the large sampling error variance which was about 134 percent of the total variance (Table 12). Therefore, a specific trend for decrease in additive genetic variance over generations could not be supported in this experiment even though the loss of additive genetic variance is the rule when genetic selection for a trait is Inade. The heritability estimate for high body weight and low body Weight selection are not statistically significant (P > .05). 78 Table 12. Analysis of variance of heritability estimates for four— week body weights from full-sibs correlation. Z Variance Sources DF SS MS "F" contribution to mean Line x rearing condition 3 .0267 .0089 <1 7 Up vs. down selection 1 .0033 .0033 <1 Raw vs. cooked soybean 1 .0044 .0044 <1 Generation 3 .1948 .0649 1.64 59 Linear l .0050 .0050 <1 Quadratic l .1173 .1173 2.94 ERROR 9 .3568 .0396 34 79 HR High line fed with raw soybean LR Low line fed with raw soybean HC = High line fed with cooked soybean LC Low line fed with cooked soybean Heritability 0 l 2 3 Generations b bk Figure 9.--Population parameter heritability estimate of 4-week body weight for line by rearing condition and generation. 80 So, these estimates were pooled to point estimate the population parameter heritability with regard to the rearing environment (one of the objectives of this experiment). Thus, these estimates are 0.36 i; 0.09 and 0.39 i;0.09 for birds in the rearing environment containing raw or cooked soybean ration, respectively. Since from the ANOVA, the effect of rearing conditions is not statistically significant, the best point estimate for this population parameter may be taken as 0.36 i 0.09 for the four-week body size. The low standard error attests to the reliability of this point estimate. Siegel (1962) reviewed 176 heritability estimates, reported by different methods for different ages (from 6 to 12 weeks), and found that most of these estimates fall within an interquartile range of 0.29 to 0.54 with the median at 0.41. The present estimated herita- bility for growth to 4 weeks of age of birds receiving diets containing either raw or cooked soybean is quite in agreement with the estimates reviewed in the literature and may be considered as a moderate herita- bility. Therefore, this lends support to the hypothesis that despite the presence of a growth depressant factor in the ration, mass selec- tion would move the mean forward due to the moderate heritability of the trait under selection. Referring to Table 10, it may be seen that the observed responses in the high line for the first and fourth generations are positive ‘values whereas values for the second and third generations are negative. However, the observed responses for all the generations in the low line are positive values. The reasons for a linear response in the low line and a non-linear response in the high line have been discussed pre- Viously. The interesting fact in Table 13 is the complete agreement 81 .mmaHH o3u onu How Hmfiuamuowmwp soauooaom w>flumasaso suHB omoommou o>Humfisaso waflvfi>fip an vmcwmuno mm: mafiafinmufiuoz vmuwamom« om.o om.o wHH he c.aw H.mo H.6m m.om mm 8.0 .i 1 n- 3: u- was I we cm . o E u- I 6.8 l. a .S I Ho 0 om.o Hm.o mwa om H.00H w.mm H.0m m.~m m Aumumamumm «zufiafinmufiuo: Amcv moafia mwcwa wouoomxm vo>uomno vouoomxm vo>ummno .smo coaumwsaoav pouaamom o3u pom osu you w: Hmwusmummmau omaommmu .mwfiv sowuooamm .mmap coauooaom coauooaom .820 .aoo ocwazoq mafiaswwm .cowumumaom >9 mafia pom uLmHmB zpon xmo3IH50m pow quHHnmuHuo: woufiammu wow omcoammu .Hmfiuaoummmav oofiuooamm .MH magma 82 between the realized (th) and the estimated population parameter (hEZ) for four-week body size in the same rearing condition. The independent realized heritability estimates for the high body weight and low body weight selection for the first and fourth generation are 0.17 (i.e., 14 42 22 25 83)’ 0.42 (i.e., 100), 0.39 (i.e., 56) and .38 (i.e., 650’ respectively. Correlated Response A concomittant change for an unselected trait which accompanies selection for another trait is considered as a correlated response. In this experiment the fourdweek body size of chicks fed with the growing ration containing cooked soybeans is a correlated response to the selection for body size of the chicks fed with the growing ration con- taining raw soybean. Referring to Table 8, it is evident that although there was a quadratic response for the selected trait in the high line, the linear decline of the correlated trait occurred after the second generation. However, the linear decline of the body size at fourdweeks of age in the selected low line on the ration containing raw soybean was accompanied by a linear decrease of the correlated body size of low line chicks receiving the ration containing cooked soybean. The asymmetry of correlated response in the high line was not expected even though it is not unusual in selection experiments. Bohren 55§.§l. (1966) studied the conditions necessary for the development of asymmetrical correlated response using simulated selection in a simple genetic model. They found that loci influencing two traits indepen- denxfly'contributed little to asymmetry while pleiotropic loci producing negative covariance with frequencies other than .5 contributed most. 83 The contributions from unequal numbers of loci or unequal effects of loci were of less importance. Also the rate of development of asymmetry was inversely related to the number of genes influencing the traits. Since in the present study, asymmetry was evident in the early genera- tions, the number of genes contributing negatively to the covariance of the two traits may be relatively very few. Nordskog and Festing (1962) observed the asymmetry of correlated -‘-_1.—. '1. in, response for body weight during selection for egg weight, and vice versa, in Leghorns in the very early generations of selection. [“ .- Correlated Response in Reproductive Fitness In Table 14, the means adjusted for line effect for fertility percentage, hatchability percentage and any hatching abnormalities constituting the components of reproductive fitness have been presented on a within line by generation. There was a significant difference (P < .01) between percent fertility for the two lines. However, over generations there was an increase trend in the percent fertility in either line (Figure 10). The mean percent fertility for the low line was 68.5 as against 60.0 for the high line. There was also an increase trend in hatchability percentage for both lines. The mean hatch— ability percentage over four generations for the high and the low line was 73.2 and 71.7, respectively; the difference was not significant (P > .05). Similarly there was no significant difference between hatching abnormalities in the high and low line. Nordskog and FeSting (1962) observed a correlated response in fertility percentage While.selecting for body size at 20 weeks of age in Fayoumi and Leghorn chickens. They concluded that large body size disfavors the 84 Table 14. Mean percentage of fertility, hatchability and hatching abnor- malities (adjusted for line effect) as correlated response to selection of body size at four weeks of age for the high and lowlines over generations. Hatching Fertility Hatchability Abnormalities (Z) (Z) (7») Gen. Highline Lowline Highline Lowline Highline Lowline 1 47.0 57.0 65.0 67.0 19.0 21.0 2 68.0 56.0 74.0 46.0 17.0 13.0 3 58.0 80.0 61.0 89.0 15.0 13.0 4 67.0 81.0 93.0 85.0 6.0 8.0 Aver. 60.0 68.5** 73.2 71.7 14.2 13.5 Regression coefficient for line eff. and overall mean for each generation Fertility Hatchability Hatch abnormalities Regression Regression Regression Overall coefficient Overall coefficient Overall coefficient Gen. mean line mean line mean line 1 0.52 -.05** 0.65 -.01 0.20 -.01 2 0.62 0.06** 0.60 0.14 0.15 0.02 3 0.69 -.ll** 0.73 -.12 0.14 0.01 4 0.74 0.07 0.89 0.04 0.07 -.01 **P < .01 1001 90‘ 801 70 O‘ 0 Percentage u: 0 40- 30* 20 10 85 ’0 High line hatchability I I I “ Low line hatchability I . , “‘Low line fertility High line fertility l 2 3 4 Generations Figure lO.--Correlated response of fertility and hatchability Percentages to the selection of 4—week body weight for the high and low line offspring . 86 fertility index in either of these breeds of chickens. The same is true in this experiment with respect to the fertility index. However, due to an increase trend in the fertility index over generations, it may be concluded that although the ration containing raw soybean depressed the growth rate of chicks fed with the ration, it did not affect the overall fitness index. Genetic Correlation The genetic correlation between two traits within line (four- week body size for chicks fed with raw and cooked soybean ration) has been computed according to the method discussed and described by Robertson (1959), Yamada (1962) and Magee (1969). On within line by generation the analysis of variance assuming mixed model (Yamada-- the two-way factorial analysis--sires as random factor and rearing condition as fixed factor) is presented from the first to the fourth generation in Tables 15a, b, c and d, respectively. The estimates of variance components for between-sires and sire x rearing condition interaction were computed according to Henderson (1953) Method I and are shown in Table 16 for the respective generations. Since negative estimates were obtained in some generations, the genetic correlation on within generation was not possible to calculate. So, these variance components were pooled over generation. The 6(bi), the squared dif- .ference between the standard deviations of between-sires within re— Spective rearing environment, was pooled over generations. Thus, the genetic correlation was .78 :_.13 and .86 i .09 for the high and the low JJJIe, respectively (Table 18). Since the estimates of variance com- Ixhnents were pooled, the directionality of the genetic correlation could Table 15. ANOVA between rearing environments within line for 4-week 87 body weight by generations. (a) Generation 1 Sources df SS MS E(MS) High line Rearing environment 1 341529.7 341529.7 Sires 4 44150.8 11037.7 owz + 13.8 OD2 + 39.5 82 Dams/sires 19 65985.0 3472.9 ow2 + 6.8 0D2 Sires x environment 4 11869.2 2967.3 ow2 + 17.4 USE2 Error 171 573391.4 3353.2 0 2 Total 199 1182173.0 Low line Rearing environment 1 883052.0 883052.0 Sires 3 114925.8 38308.6 owz + 17.4 OD2 + 64.4 52 2 2 Dams/sires 19 96277.5 5067.2 ow + 10.5 CD Sires x environment 3 25578.4 8509.5 ow2 + 30.9 OSEZ Error 243 660641.5 2718.7 0 2 Total 269 l916282.4 Table 15. 88 body weight by generations. (b) Generation 2 ANOVA between rearing environments within line for 4-week Sources df SS MS E(MS) High line Rearing environment 1 656427.5 656427.5 Sires 5 18607.9 3721.6 owz + 10.8 OD2 + 37.0082 Dams/sire 22 166050.4 7547.7 CW2 + 7.2 OD Sires x rearing 2 2 environment 5 34957.3 6991.5 Ow + 16.7 USE Error 191 589367.0 3085.7 owz Total 224 1981720.9 Low line Rearing environment 1 2825123.1 2825123.1 Sires 4 29159.5 7289.9 owz + 14.7 oDz + 70.8082 2 2 Dam/sire 22 153074.5 6957.9 ow + 13.0 CD Sires x rearing 2 2 environment 4 22559.3 5639.8 ow + 34.9 USE Error 329 1024300.6 3113.3 owz Total 360 4952878.3 Table 15. 89 body weight by generations. (c) Generation 3 ANOVA between rearing environments within line for 4—week Sources df SS MS E(MS) High line Rearing environment 1 150500.2 150500.2 Sires 5 11749l.6 23498.3 0W2 + 9.2 ODZ + 39.6032 Dams/sires 20 252506.0 12625.3 0W2 + 10.3 ODZ Sires x rearing 2 2 environment 5 11861.3 2372.3 ow + 21.8 USE Error 234 748893.2 3200.4 0 2 Total 265 l936384.6 Low line Rearing environment 1 1457327.3 1457327.3 Sires 5 '109497.0 21899.4 ow2 + 10.5 ow2 + 66.1082 Dams/sires 33 216396.5 6557.5 ow2 + 10.3 61)2 Sires x rearing 2 2 environment 5 12860.1 2572.0 ow + 31.7 USE Error 358 1097929.9 3066.8 6x2 Total 402 3397458.7 Table 15. 90 body weight by generations. (d) Generation 4 ANOVA between rearing environments within lines for four—week Sources df SS MS E(MS) High line Rearing environment 1 107668.6 107668. Sires 4 20220.8 5055. ow2 + 9.5 OD2 + 18.1032 2 2 Dams/sires 9 21766.3 2418. GW + 6.3 CD Sires x rearing 2 2 environment 4 13823.7 3455. Ow + 7.8 OSE Error 74 236853.2 3200. ow2 Total 92 472277.9 Low line Rearing environment 1 296602.2 296602. Sires 4 13409.3 3352. 0W2 + 7.3 0D2 + 33.3632 2 2 Dams/sires 18 65150.5 3563. ow + 6.1 CD Sires x rearing 2 2 environment 4 13300.0 3325. GW + 15.9 USE Error 151 340841.9 2257. ow2 Total 178 1376451.1 91 I. . l . .‘ . alt—HI .uxou on» cw wonfiuomop mum mnonuoa HmGOfiumusmaoo was maOHumuozx o.~HH N.ooH o.o~o H.~HH m.aa m.onH owooo>o ooaooo m.o~a- o.ro- o.- H.N~ ~.ao H.o o N.Nmm o.amo o.~Ho o.~ o.mH- m.oa~ m m.oa o.oo o.raH o.~o o.~a m.~- N m.ao~ o.NoH o.wooa o.oom o.ara o.Noq H ooafl 3o; m.wo~ N.Hmo m.qq~ 8.68 o.Hm H.m~H owooo>o ooaooo o.Hm~ o.~om o.qam o.m n.~m o.NoH o o.mrw o.ma¢ o.maa 0.65 o.rm- o.oom m o.oNH- o.amo o.~o~ m.~m o.mm~ a.moa- N 0.5mm o.Hoo o.NrH “.ma N.~N- o.rwa H ooHH swam omrmooo onmmm «Aromm Aoovm moor Now oooooooooo ooooo «Aoooav oowoz rnmoofiv oooao» .muocuoe msofinm> an aofiumaonuoo ofiuocow mo GOHumusmEoo pom monocomaoo woumafiumm .oH manna 92 not be ascertained over generations. Therefore, Robertson's (1959) method with assumption of a fixed model (sires assumed as fixed factor?) was used to compute the genetic correlation within generation. The same tables of analysis of variance on a within line by generation were used to compute the genetic correlation from mean square esti- mates instead of variance components. Thus, the genetic correlation from 1 to 4 generations in the high line was 1.11, -.72, 1.08, and .76 and in the low line the same estimate was .72, .24, 1.05 and .01, respectively. However, the weighted average was computed to estimate the population parameter by pooling the sum of squares and respective degrees of freedom for respective sources of variation (sires, sire x rearing environment and error). The pooled mean squares for the re- spective sources of variations were 11137, 4885 and 3207 for the high line and 16687, 4640 and 2887 for the low line. Thus, the best esti- mate of the genetic correlation was .65 i .20 and .77 1:14 for the high and the low line, respectively (When these mean squares are adjusted for 0(bi) value as discussed by Yamada (1962), the estimated genetic correlation closely agrees to that of Yamada.). The estimates within generation are plotted in Figure 11 which indicates no specific trend of the genetic covariance over generations. While estimating the genetic correlation by variance and co- variance technique (Magee, 1969, Falconer, 1960) the covariances be- tween unadjusted sire-progeny means (Table 17) on different rearing environments were pooled over generations. Thus, the genetic correla- tions obtained for the high and the low line were .92 i .05 and .48 i .27, respectively. By this method the estimate in the high line was above the value obtained by Yamada (1962) and, similarly, in the low Genetic correlation 93 2.0‘ 1.54 1.03 0.5- Generations -1.0 -l.5' -2.0- Figure ll.--Estimated genetic correlation between selected and unselected traits within generation by lines (Robertson's method). 94 Table 17. Sire-progeny unadjusted means for body weight in grams at four weeks of age by lines, generations and rearing environment. High Line Low Line Raw soybean Cooked soybean Raw soybean Cooked soybean No. of No. of No. of No. of Gen. off./sire Mean off./sire Mean off./sire Mean off./sire Mean 1 18 429 28 544 31 372 21 467 15 418 15 473 46 331 51 482 31 393 22 497 54 399 32 510 23 434 6 531 18 337 17 487 2 17 415 18 574 46 298 40 497 14 377 19 511 16 332 21 516 24 387 19 586 36 310 28 494 29 416 25 528 40 290 40 522 17 400 20 551 42 331 52 522 3 33 346 21 446 40 280 24 412 43 352 50 462 37 246 20 376 5 435 7 526 27 257 17 414 38 384 36 504 51 275 45 391 1 442 4 487 42 311 47 441 17 392 11 479 30 288 23 444 4 9 433 5 470 24 258 27 382 15 402 8 487 3 284 5 375 10 365 7 497 9 247 11 409 8 429 3 565 33 239 25 389 10 403 18 479 16 261 26 416 11$“---ng 4.- ,.. v '- 95 Table 18. Estimated pooled genetic correlation over generations between four-week body weight expressed in two dietary environments ”an (for the high and low lines). ' Lines Models Methods High Low assumed [ Robertson (1959) 0.65 i 0.20 0.77 :_0.14 *Fixed model Yamada (1962) 0.78 i 0.13 0.86 i 0.09 Mixed model Magee (1969) 0.92 i 0.05 0.48 i 0.27 Variance & covariance *6(bi) has not been adjusted (see text). 96 line, it was far below the estimate computed by Yamada's method. Probably, the sampling error may be the cause of this disagreement be- tween two-way ANOVA and variance-covariance techniques (based on unad- justed sire-progeny means). The magnitude of the standard error (computed by the formula described by Falconer, 1960), indicates that the genetic correlation, regardless of method used for computation, significantly deviated from 1. 176-3 0“; “ J Robertson (1959) pointed out that a genetic correlation of .8 with a standard error of .2 may be taken as a significant deviation from 1 in a situation where the same genotype is allowed to express itself in two different environments. Since the expected genetic correlation for a situation where the same genotype expressed in two or more environ- ments is 1, any significant deviation therefrom may be attributed to the genotype-environment interaction (Robertson, 1959). Sex Effect and Sex by Rearing Environment Interaction The least square estimate for sex effect as a deviation from the mean had a range from 14.3 to 21.7 grams over four generations and these estimates were highly significant (P < .01) (Table 7). Thus, a significant difference between 4-week body size for male and female, when averaged over two rearing environments as a deviation from zero, had the range of 28.5 to 43.5 grams. Similarly, from the same table the difference between two rearing conditions when averaged over sexes varied from 109.6 to 168.4 grams over four generations. The effect of sex by rearing environment interaction was consistently a negative 97 estimate varying from -6.2 to -13.2 and was highly significant (P < .01) in all of the four generations (Table 7). Therefore, this negative interaction effect reflects on the four-week body size of the males and females as an unequal difference between sexes reared in dietary environ- ments where they received raw soybean as compared to the difference in body size between sexes in the other rearing environment. In other words, the males fed with raw soybean ration grew less in body size relative to the contemporary females reared in same dietary environ- ment. However this was not true for the rearing environment where birds were fed cooked soybean. This is evident from a comparison of the magnitude of the sex difference in four week body size of chicks fed raw soybean in the high line from the first to the fourth genera- tion where the values were 30, 6, 22 and 12 grams, respectively, with that of the chicks fed with cooked soybean ration where the respective sex difference in body size during the same generations was 56, 51, 44 and 48 grams, respectively (Table 8). Thus, a decline of the sex dif- ference (Figure 12) over the four generations on the raw soybean dietary treatment and an approximate steady state situation in sex difference for individuals fed with the cooked soybean ration possibly could be due to the existence of differential degrees of tolerance between sexes when receiving the diet containing raw soybean either due to a sex- linked or sex-influenced effect. In Table 19 the total number of chicks hatched, total number of chicks surviving, by sexes, at 4 weeks of age, the percent mortality and the survival sex-ratio at the selection age are shown on within generation-rearing condition by line. Many questions could be asked while looking at this table. Two of the important questions are: Absolute difference of 4—week body weight (gms) between male and female Absolute difference of 4-week body weight (gms) between male and female 98 \ 100‘ \ . \ \ \/. 80* ‘\ . \ \ \ \ 60‘ \ o 40‘ \ Q 4 20* lst Generation 2nd Generation 0 , - T r M F M F 1004 O \ \ 80‘ \ 60~ 40‘ \ a 4th Generation 3rd Generation 20 M F ‘M f— ' Figure 12.--The effect of sex by rearing environment as an inter- action deviation from overall constant within a generation basis. 99 Hm.0 00.0 0.0 m.0 0.05 0.00 0.00 m.~0 Hma 00 00m voxoou 00.0 00.0 0.0 H.n 0.0m 0.0m N.¢n m.Hm 00H NHH 00m 3mm owmuo>¢ 55.0 mo.H n.0H 0.0H mm as om Hm OHH mo aoo ooxoou 00.0 00.0 0.0a 0.5H 00 0m am am 00H m0 00m 3mm c 00.0 m0.0 0.0 n.0 00 N0 00 00 “NH 00H mom 003000 00.0 H0.0 0.0 H.~ 0HH 00H 00 mm mew 00H 00m 3mm m aw.o 0N.H «.6 m.o am or so mm 0rd HHH aoo ooxoou 00.0 w~.H N.0 0.0 00H mm mm 00 NOH 00H how 3mm N mm.0 m0.0 0.0 0.0 00 mm 00 me mad m0 00m voxooo 00.0 mH.H 0.0 n.H 05 mm mm 00 00H mHH 00m 3mm H 30A cwfim 304 awwm mamaom mam: onEmm onz 3OA swam uamsooufi>co .aou BOA swam wawumom AonEmm " mamzv oaumu xmm vo>H>u5m Hm>H>MSm mxowno 00 .oz .moc«H 0am coaumuooow .ucoaaoufi>ao wowumou 00 oaumu xom Hm>H>usm 0am owmucmouon 0ufifimuuoa .mww mo mxoo3 snow on on>H>usm amass: .0osoumn mxoano mo nonasz .0H manna 100 1. Is there any differential mortality with respect to rearing conditions that may affect the sub-class mean performance? 2. Does the expected sex ratio differ from line to line? [The possible existence of a differential degree of tolerance by sex due to raw soybean diet has already been discussed. Since the high line chicks grow better than the low line chicks, they might be expected to better tolerate the depressing effect of raw soybean than would the low line chicks. There- fore, a deviation in sex ratio.from 1 : 1 due to loss of more male chicks in either lines (because of l. overactivity of the vital organs such as the pancreas in high line males, or 2. threshold limit for the survival of the low line males) may be the other possibility to account for the unequal subclass mean performance.] Both of these questions could be answered by appropriately testing the hypotheses by the xg-test. The xz-value for the differential mortality with respect to rearing condition was .28 indicating no significant difference (P > .05) between lines in livability during 4 weeks of growth. Similarly, for the expected sex ratio between lines, the 52- value (1.01) was not significant (P > .05). This indicates that there exists no differential sex survival in the two rearing conditions. Therefore, it may be concluded that any existing relative degree of intolerance to depressing effect for body size in male chicks fed with raw soybean did not cause a differential sex mortality. The average sex-ratios (male : female) in the high and low lines were .95 and .86 for the dietary environment where birds were fed raw soybean and the 101 ratios for the environment where birds received rations containing cooked soybean were .97 and .81, respectively, with a random deviation from generation to generation. Sex—ratios were not significantly dif- ferent in the same line on different rations. Therefore the depres- sant factors present in raw soybean are not lethal to either sex. Semi-lethality may be postulated due to the following reason. Since a male chicken is homogametic for the sex chromosome, the relative degree of intolerance to the depression effect when reared in the raw soybean ration environment may be higher as compared to a female which is heterogametic for sex chromosomes. This would be true if there are specific genes on the x-chromosome that are responsible for the expres- sion of body size only when the chicks are fed with the raw soybean ration. Diallel crossing may unfold the answer to this postulation. Brunson 35 a1. (1956) observed a sex-linked gene effect on lO—week body weights of chickens. Siegel (1962), in discussing the gene action for a quantitative trait like 8-week breast angle, confirmed the sex-linked gene effect as observed by Brunson 35 31. (1956) for 10-week breast angle. Expected Phenotypic Correlation Between the Selected and Correlated Traits The expected phenotypic correlations within lines estimated by the following methods are presented in Table 20. 1. By the direct product moment correlation computed by use of the sire progeny unadjusted means (Table 17): Cov p r __ 192 9192 cpl 092 102 2. As a function of heritability and genetic correlation r = r 2 2 p1P2 G162 Vh1 \Ihz 3. By the path coefficient method 2 rPorPQc =<\(N+l)gh :>A 4 + h [(NH)n-2] rQ The heritability estimate for either line has been taken as .36 1;.09 in all the above methods and the point estimate for genetic correlation A (rG) has been assumed to be .78 :_.13 for the high line and .86 i .09 for the low line (Yamada, 1962). Table 20. Expected phenotypic correlations by various methods for the selected and correlated traits within line. Methods Lines 1 2 3 High line 0.35 0.28 0.54 Low line 0.81 0.31 0.64 The estimate from the path diagram method (Method 3) is comparable to that from the direct product moment correlation method (Method 1) for the low line but not for the high line. However, by the second method the estimate for the low line is far below that computed for the same line by either of the other two methods. This may be due to sampling error. APPLICATION Genetic parameters such as heritability, genetic correlation, etc. when estimated from a given population at a given time may or may not be the same for another population at another time. Therefore, care should be taken when applying these parameters. In a short term selection program such as this experiment, the reliability of the point estimate for any given parameter must be based on adequate degrees of freedom to avoid overlapping in confidence in- terval. However, in any experiment sample size is assumed as the representative sample drawn from the normal population. Even when the utmost care is taken in the sampling procedure, the sampling error is difficult to overcome. Therefore, the reliability for any genetic parameter depends on a basis of trend. This experiment was designed to separate two lines of chickens on the basis of their 4-week body size when both lines were receiving a ration containing raw soybean. Since raw soybean contains some growth-depressant factors, the nature of which is not completely known, the success in selecting for large body size and small body size was assumed to be the over-coming of the effect of the depressant factors in the high line and intolerance to these factors in the low line. In the present study, the heritability estimate (.36 i .09) for 4-week body size based on the full-sib correlation for mixed sex, and including all birds, that is, those receiving the ration containing 103 104 raw soybeans and those receiving the ration containing cooked soybean, was in agreement with those reviewed by Siegel (1962). This estimate may be taken as a moderate one and therefore a mass selection accom- panied by selection-aids probably would improve the population means at a faster rate than that which was obtained by only mass selection. Further, the equal heritability estimate in either rearing environment would mean the mass selection practiced on any population could be equally efficient for direct and correlated response if the genetic correlation between two traits is nearly 1. To verify the above hypothesis some of the third generation off- springs for both the high and low line reared in each of the two environ- ments (raw or cooked soybean ration) were selected on their own pheno- type (4-week body size) to become parents for the fourth generation offsprings. The fourth generation offsprings being fed with raw as well as cooked soybean ration were subjected to analysis of variance by the usual least squares procedure; thus involving the mean performances of eight subclass groups. The least squares estimates for main and first order interaction effects as a deviation from the mean are shown in Table 21. The adjusted subclass means are presented in Table 22 and Figure 13. The analysis of variance for the subclass means is given in Table 23. From the analysis of variance, it is seen that the three main factors, i.e. rearing condition, line and sex, were highly significant (P < .01) while the factor, the selection of parents, was not significant (P > .05). This means that the type of parental selection may have no significant effect on the subclass means. 105 Table 21. Least squares estimates for main factors and first order interaction effects with S.E. for the different sources of variation for the subclass means for four—week body weight of fourth generation chicks reared in two dietary environ- ments. Overall constant: 383.61 Rearing environment: -51.39 i 2.44 Line: 52.69 i 2.44 Sex: 20.60 :_2.44 Parental selections: 2.41 i 2.44 Rearing environment x line: 14.58 :_2.44 Rearing environment x sex: -6.31 :_2.44 Rearing environment x parental selection: -6.13 :_2.44 Line x parental selection: 8.07 i 2.44 Line x sex: 0.77 :_2.44 106 Table 22. Adjusted subclass progeny means (by lines and selection of parents) in two rearing environments (fourth generation). Rearing environment Raw soybean ration Cooked soybean ration Subclass by selection 3rd gen. 4th gen. 3rd gen. 4th gen. of parents mean mean mean mean High line - parental selection from raw 384 404 -—- 490 soybean fed chicks Low line - parental selection from raw 283 253 --- 398 soybean fed chicks High line - parental selection from cooked --- 395 495 457 soybean fed chicks Low line - parental selection from cooked --- 278 418 396 soybean fed chicks 107 Progeny fed with ration containing cooked soybean Progeny fed with ration containing raw soybean 500 “181‘ \\ .\\\ . _‘ \ \ ‘ \ w: A450 \\\\.\\ a :§:‘\:\‘\y " \N 8 \ . m 0400 \Q\ 0 Low (D x o w 3 H350 : o ‘H u m 2300 m o E h s 0 DD 3250 0. Low 700 Parents selected Parents selected from raw soybean from cooked soybean fed individuals fed individuals PARENTS Figure l3.——Geometric presentation of the performances of the high and low line progeny sired by third generation parents being reared on ration containing raw or cooked soybean. 107 Progeny fed with ration containing cooked soybean Progeny fed with ration containing raw soybean 500 “159 <\ \. \.\\\\\\\\\; A450 \ \ 8 RN 8° \ \\ :10 \ \ £400 \ 0 Low In x o o 3 p350 : o ‘44 u m U) $300 0 E 0 c o O!) 3250 0 Low 9nn Parents selected Parents selected from raw soybean from cooked soybean fed individuals fed individuals PARENTS Figure 13.——Geometric presentation of the performances of the high and low line progeny sired by third generation parents being reared on ration containing raw or cooked soybean. Table 23. 108 Analysis of variance of four-week body size for Generation 4 whose parents were selected from environments where birds received rations containing either raw or cooked soybean. Sources df SS MS "F" Rearing environment 1 l357430.8 l357430.8 443.02** Line 1 1426706.4 1426706.4 465.62** Sex 1 218036.5 218036.5 71.15** Parents (selection) 1 3001.1 3001.1 0.98 Rearing environment line 1 109223.2 109223.2 35.65** Rearing environment sex 1 20499.5 20499.5 6.69** Rearing environment parent 1 19339.1 19339.1 6.31** Line x sex 1 308.1 308.1 1 Sex x parents 1 8987.6 8987.6 2.9 Rearing environment line x sex 1 38.6 38.6 1 Rearing environment line x parents 1 781.9 781.9 1 Rearing environment sex x parents 1 606.8 606.8 1 Line x sex x parents 1 151.1 151.1 1 Rearing environment x line x parents x sex 1 4027.6 4027.6 1.31 Error 557 1706684.0 3064.1 Total 572 5064265.l *P < .01 109 The first order interactions due to the type of selection of parents with rearing environment and line, which were highly significant (P < .01), need a further discussion. 1. Parental Selection x line interaction: In the high line the significant least squares estimate (P < .01) for this inter- action effect was 8.1 gms as a deviation from the mean for the offsprings of the selected parent when grown in the environ- ment where they received the ration containing raw soybean. Thus the difference between the high and low lines was greater when the selected birds had been raised on raw soybean than when they had been raised on cooked soybean. 2. Interaction effect for rearing environment x parental selec- tigg; The offsprings of the parents which had been reared and selected from one environment showed an interaction effect of -6.1 gms when subjected to the parental rearing environment; whereas, when subjected to the opposite environ- ment the interaction effect was 6.1 gms. Referring to Table 22, for the high line the subclass mean per- formances at 4 weeks of age for the offsprings (sired by a group of parents reared and selected from a growing ration containing raw soybean) receiving the ration containing raw soybean and for those receiving the ration containing cooked soybean, were 404 and 490 grams, respectively. The offsprings produced from the parents reared and selected from the growing ration environment containing cooked soybean weighed 395 and 457 grams on the two respective rations. This is partly in agreement with Falconer's (1959) conclusion regarding the selection experiment on body size in mice from a high and low plane of nutrition. He showed that 110 mice reared on a low plane of nutrition and selected for growth rate responded in a distinctly different manner to high and low plane diets than did mice reared and selected on a high plane diet. The strain selected on the low plane performed well on either low plane or high plane nutrition (comparable here to the strain performing well in either raw or cooked soybean ration in this experiment); whereas, the high plane selected lines only did well on high plane diets (not comparable to this experiment, where a decline of 38 grams from third to fourth generation was observed). Since in the present experiment selection was practiced on the third generation offsprings, it is reasonable to accept that a decline in performance of the offsprings sired by the parents reared and selected from the cooked soybean fed line is not due to any specific cause other than sampling error. A few more generations of selection might confirm in full the conclusion by Falconer (1959). Genetic Correlation In the high line the genetic correlation obtained by different methods had a range from .65 to .92 and was significantly different from the expected genetic correlation of 1. The range of genetic correlation (from .48 to .86) in the low line was very similar to that obtained for the high line. Genetic correlation between two traits and the respec- tive heritability estimates are needed to compute a correlated response for a trait from the direct selection of another trait. CONCLUSIONS Results of this study indicate that: There is no difference in the heritability estimates for the body size at four weeks of age for the chicks receiving the raw or cooked soybean ration. Due to moderate heritability estimates, mass selection for tolerance to raw soybean in the diet of chicks from hatching to four weeks of age should be aided by family selection or other kinds of selection- aids to move the population means faster. The selection of individuals from the stressed dietary environment (e.g. ration containing raw soybean) is equally effective in in- fluencing the performance of their progeny either in stressed or "normal" dietary environments whereas seleCtion of individuals from the "normal" dietray environment (e.g. ration containing cooked soybean) may not be as effective when their progeny are raised in the stressed dietary environment. The genetic correlation between fourdweek body Size in the two rearing environments in this experiment was significantly different from 1. Therefore, the prediction of the carryover of the progress made in one rearing environment to another environment will depend upon the magnitude of the genetic correlation of the traits. lll SUMMARY The selection experiment was designed to establish two divergent lines of chicks based on their body size at four weeks of age, when fed with a growing ration containing raw soybean. The unselected trait, the body size at the same age for chicks of the same genetic group reared in an environment where they received cooked rather than raw soy- bean in their ration, was considered as a correlated response. Over four generations and for a period of four years, from 1965 to 1968 in- clusive, a record of 1997 chicks produced from 42 males and 204 females of the White Plymouth Rock breed of chickens was included in the data for an analysis. In the high line, observations were on 784 chicks produced by 22 sires and 92 dams; whereas, in the low line the record had 1213 observations on chicks obtained from 20 sires and 112 dams. Since the offsprings from the same genetic group by lines were permitted to express their body size at 4 weeks of age in two dietary rearing environments, the hierarchical situation of the experimental design (dams being nested under sire by year) became a cross classified mixed model. Due to unequal frequency per treatment combinations, the least squares procedure was considered the best one for analyzing the data. The data were analyzed on a mixed sex basis by considering sex as a separate factor being cross classified with sire and dam. 112 113 The results were as follows: Genotype (line) x rearing environment interaction: The significant least squares estimate (P < .01) for this interaction had a range from 6.04 to 13.98 without any specific trend over four generations. The independent line effect had a linear increase trend. The re— gression of this line effect on generations was 14.2 gm per genera- tion. Similarly, the independent effect due to rearing environment showed a linear decline from the second generation to the fourth generation and the rearing environmental effect computed as a re- gression on generation was -14.7 gm per generation. Thus, the small change (-.5 gm) per generation in the high line was the result of a balancing effect of the line (genotype) due to a positive selection for body size and a dietary depression influenced by the rearing environment. These two effects in the low line were accumulated per generation (i.e., [-14.2] + {-14.7} = -28.9) and therefore had a linear change from one generation to the next. The variance com- ponent due to genotype x environment interaction on within line by generation was estimated by twodway factorial model as described by Robertson (1959) and extended by Yamada (1962). This was estimated so that the genetic correlation between the body size at four weeks of age expressed in two dietary rearing environments could be com- puted. The genetic correlations for the high and low line computed by this method by pooling the respective variance components over four generations were .78 :_.13 and .86 i .09, respectively. Applying the variance and covariance technique for computation of genetic correlation the estimates for the high and low line were .92 i .05 and '48.i.-27, respectively. Regardless of the method 114 of computation used, these estimates were significantly different from the expected genetic correlation of l. Correlated response: In the high line the concomittant change for the body size at four weeks of age in chicks fed with cooked soybean had a declining trend over four generations thus causing an asymmetry with respect to increase in body size for the chicks fed with the growing ration containing raw soybean. The correlated response for the unselected trait in the low line was not discordant with the linear decline of the selected trait. The correlated response with regard to reproductive fitness had an increase trend over generations. The mean fertility per- centage adjusted for line effect for the high line was 60.0 vs 78.5 for the low line and the two lines were significantly different for this character (P < .01). The mean hatchability percentages over four generations for the high and low lines were 73.2 and 71.7, respectively, and showed random deviation from generation to genera- tion. However, the mean of hatching abnormalities over generations had a declining trend in both lines and was not significantly dif- ferent in the two lines. By a hierarchical model on a within rearing environment, line and generation, the heritability for 4-week body size was estimated from full-sib intra-class correlation. The heritability estimate on a within generation by line had a range from .08 to .73 with a standard error from .18 to .32 in the high line for the selected trait (body size in raw soybean dietary environment). A similar trend for the heritability estimates for the selected trait in the low line as well as for the correlated trait in the high and low line, 115 respectively, was observed. When these estimates were pooled over four generations within rearing environment, the population parameter was .36 t .09 for both the selected and correlated traits. Due to the equal moderate heritability for body size at four weeks of age, regardless of dietary rearing environments, mass selection from either environment would be equally efficient in moving the pOpula- tion mean forward, depending upon the magnitude of the genetic cor- relation of the two traits. The realized heritabilities computed by the ratio of actual responses over cumulative selection differentials of the two lines on a within generation (first and fourth generation only) were in agreement with the population parameter. They were .31 and .39 for the first and fourth generation, respectively, with an average of .35. Sex by rearing environment interaction: The significant least squares estimate (P < .01) had a range from -5.45 to -11.21 with random changes from one generation to the next. The negative estimate for this interaction effect implies that the males did not grow as well as the females in the dietary rearing environment con- taining raw soybean. Sex by line interaction: The effect due to the sex by line inter— action was not significant in any of the generations. This indi- cates that the effect of sex on body size in either line was inde— pendent of the line effect or vice versa. Type of parental selection x rearing environment: The least square estimate for the parental selection x rearing environment inter- action was highly significant (P < .01). Progeny, from parents 116 which had been selected on the basis of performance when receiving a ration containing raw soybean, did not perform as well when re- ceiving this ration as did their sibs which received a ration con- taining cooked soybean. The difference between these two subclass means was 12.2 gms in favor of the progeny receiving the ration containing cooked soybean. Exactly opposite was the situation for the performance of the progeny produced from the parents which had been selected on the basis of performance when receiving a ration containing cooked soybean. These results were postulated to be due to the differential degree of tolerance of males and females to the raw soybean ration. So, a possible mating program in which males selected from the cooked soybean ration environment would be mated with females reared and selected from the dietary environment con- taining raw soybean so as to produce a best nicking effect might be suggested. Parental selection on rearing environment x line interaction: Progeny from the high line parents which had been reared and selected from the environment where they received a ration containing raw soybean were significantly heavier (P < .01) by 21.0 gms than progeny from the low line parents from the same dietary environment. This indicates that the selection for high and low performance in this specific dietary environment, i.e., ration containing raw soybean, was effective. In the high line, selected individuals reared in a stressed dietary environment (raw soybean ration) produced offspring that performed equally well in stressed and "normal" rearing environment (cooked soybean ration); whereas, offspring of selected individuals reared 117 in a "normal" rearing environment did not perform as well in either rearing environment as did the chicks from the stressed dietary environment high line stock. The sampling error may be the cause for this type of unexpected result. The phenotypic correlation between the selected and correlated traits was .35 in the high line and .81 in the low line, estimated by product moment correlation technique. Using the path diagram method as per Lush (1948), the respective phenotypic correlation coefficients were .54 and .64 in the high and low line. LITERATURE CITED LITERATURE CITED Abbott, U. K. and V. S. Admundson, 1962. Responses to selection under severe environmental stress. XII World's Poultry Congress Proc.: 30. Amer, M. F., 1965. Heritability of body weight in Fayoumi. Poultry Sci. 44:741. Asmundson, V. S. and I. M. Lerner, 1933. Inheritance of rating growth in domestic fowl. II. Genetic variations in growth of Leghorns. Poultry Sci. 12:250. Birk, Y., 1961. Purification and some properties of a highly active inhibitor of trypsin and Chymotrypsin from soybeans. Biochem. Biophys. Acta 54:378. Birk, Y., A. Bondi, B. Gastetner and I. Ishaaya, 1963. A thermostable haemolytic factor in soybeans. Nature 197:1089. Boheren, B. B., W. G. Hill and A. Robertson, 1966. Some observations on asymmetrical correlated response to selection. Genet. Res. 7:44. Booth, A. N., D. J. Robbins, W. E. Ribelin and F. DeEds, 1960. Effect of raw soybean meal and amino acids on pancreatic hypertrophy in rats. Proc. Soc. Exptl. Biol. Med., 104:681. Borchers, R., 1961. Counteraction of the growth depression of raw soy- bean oil meal by amino acid supplements in weanling rats. J. Nutr. 75:330. Brambilla, S., M. C. Nesheim and F. W. Hill, 1959. Effect of trypsin supplementation on utilization by the chick of diets containing raw soybean meal. J. Nutr. 75:13. Brunson, C. C., G. F. Godfrey and B. L. Goodman, 1956. Types of gene action in the inheritance of ten-week body weight and breast angle in broilers. Poultry Sci. 35:524. Bumgardner, H. L. and C. S. Shaffner, 1954. The heritability of body weight in turkeys. Poultry Sci. 33:601. Carte, I. F. and P. B. Siegel, 1968. Correlated heritabilities of unselected traits: A check on procedure. Poultry Sci. 47:1821. 118 119 Clayton, A., G. R. Knight, J. A. Morris and A. Robertson, 1957. An experimental check on quantitative genetics theory. 111. Correlated responses. J. Genet. 55:171. Dickerson, G. E. and H. Abplanalp, 1956. Sources of variation in per- formance of entries in the 4th, 5th and 6th California random sample egg production tests. Proc. 9th Pacific Chicken and Turkey Breeder Round Table 9:67. Dickerson, G. E., 1962. Implications of genetic-environmental inter— action in animal breeding. An. Prod. 4:47. Dickerson, G. B., 1969. Techniques and procedures in animal science research. Published by the American Society of Animal Science: 36. Enos, H. L. and R. E. Moreng, 1965. Evidence of genetic variability of lysine utilization. Poultry Sci. 44:966. Falconer, D. S., 1953. Selection for large and small size in mice. J. Genet. 51:470. Falconer, D. S., 1959. Selection of mice from growth on high and low plane of nutrition. Genet. Res. 1:91. Falconer, D. S., 1960. Introduction to quantitative genetics. The Ronald Press Company, New York. Flangan, J., 1969. Personal communication. Frattali, V. and R. F. Steiner, 1968. Soybean inhibitors. 1. Separa— tion and some properties of three inhibitors from commercial crude soybean trypsin inhibitor. Biochem. 7:521. Garlich, J. D. and M. C. Nesheim, 1966. Relationship of fractions of soybeans and a crystalline soybean trypsin inhibitor to the effects of feeding unheated soybean meal to chicks. J. Nutr. 88:100. Gartler, A., Y. Birk and A. Bondi, 1967. A comparative study of the nutritional and physiological significance of pure trypsin inhibitors and of ethanol-extracted soybean meals in chicks and rats. J. Nutr. 91:358. Gill, J. L., 1969. Personal communication. Godfrey, E. M., 1968. Ten generations of selection for lysine utiliza- tion in Japanese quail. Poultry Sci. 47:1559. Goodman, B. L., J. F. Grimes and R. J. Jaap, 1957. Repeatability esti- mates of genetic variance within two closed flock strains. Poultry Sci. 36:1121. 120 Goodman, B. L., and R. G. Jaap, 1960. Improving accuracy of heritability estimates from diallel and triallel matings in poultry. I. Eight weeks body weight in closed flock strains. Poultry Sci. 39:938. Goodman, B. L. and R. G. Jaap, 1961. Non—additive and sex—lined genetic effects on egg production in a random-bred population. Poultry Sci. 40:662. Gorrill, A. D. L. and J. W. Thomas, 1967. Body weight changes, pancreas size and enzyme activity, and proteolytic enzyme activity and protein digestion in intestinal contents from calves fed soybean and milk protein diets. J. Nutr. 92:215. Goto, E. and A. W. Nordskog, 1957. Heterosis in poultry. Poultry Sci. 38:1381. Gowe, R. S., 1956. Environment and poultry breeding problems. Poultry Sci. 35:430. Gowe, R. S., J. A. Lemay, and A. S. Johnson, 1962. Importance of genotype- environment interactions for quantitative traits involving com- mercial egg production strains and two rearing programs--restricted and full-feeding. XII World's Poultry Congress Proc.: 43. Haines, P. C. and R. L. Lyman, 1961. Relationship of pancreatic enzyme secretion to growth inhibition in rats fed soybean trypsin in- hibitor. J. Nutr. 74:445. Haldane, J. B. S., 1946. The interaction of nature and nuture. Ann. Eugenics 13:197-205. Hazel, L. N., 1943. The genetic basis for constructing selection indexes. Genetics 78:476. Henderson, C. R., 1953. Estimation of variance and covariance components. Biometrics 9:226. Hess, C. W., H. M. Edwards, Jr., and E. F. Derubnicki, 1962. Growth rate selection on a methionine deficiency diet. Poultry Sci. 41:1042. Hicks, A. F., Jr., 1958. Heritability and correlation analysis of egg weight, egg shape and egg number in chickens. Poultry Sci. 37:967. Hill, J. R., and A. W. Nordskog, 1956. Efficiency of performance testing in poultry. Poultry Sci. 35:582. Hutt, F. B., R. K. Cole and J. H. Bruckner, 1945. A test of fowls bred for resistance to lymphomatosis. Poultry Sci. 24:564. 121 Ideta, G. and P. B. Siegel, 1966. Selection of body weight at eight weeks of age. III. Realized heritability of unselected trait. Poultry Sci. 45:923. Jerome, F. N., C. R. Henderson and S. C. King, 1956. Heritabilities, gene interactions and correlations associated with certain traits of domestic fowl. Poultry Sci. 35:995. Kan, J., W. F. Krueger and J. H. Quisenberry, 1959. Non—additive gene effects of six broiler traits as studied from a series of diallel matings. Poultry Sci. 38:972. fl Kempthrone, 0., 1957. An introduction to genetic statistics. Wiley, New York. Khayambashi, H., and R. L. Lyman, 1966. Growth depression and pan- creatic and intestinal changes in rats force-fed amino acid' diets containing soybean trypsin inhibitor. J. Nutr. 89:455. Kratzer, D., 1969. Personal communication. .7 Krishan, A. and R. N. Shoffner, 1966. Sex chromosomes in the domestic fowl (Callus domesticus), turkey (Meleagris gallopavo) and the Chinese pheasant (Phasianus colchicus). Cytogenetics 5:53. Kunitz, M., 1946. Crystalline soybean trypsin inhibitor. J. Gen. Physiology 29:149. Kwong, E., R. H. Barnes, and G. Fialla, 1962. Intestinal absorption of nitrogen and methionine from processed soybean in rats. J. Nutr. 77:312. Kwong, E. and R. H. Barnes, 1963. Effect of soybean trypsin inhibitor on methionine and cystine utilization. J. Nutr. 81:392. Lamoreux, W. F. and F. B. Hutt, 1948. Genetic resistance to deficiency of riboflavin in chicks. Poultry Sci. 27:334. Lepore, D. D., 1965. Methionine and protein requirements of lines of chickens established by growth rate selection in a methionine deficient diet. Poultry Sci. 44:797. Lepkovsky, S., F. Furuta, T. Koike, N. Hasegawa, M. K. Dimick, K. Krause, and F. J. Barnes, 1965. The effect of raw soya beans upon the digestion of proteins and upon the pancreas of intact chickens and of chickens with ileostomies. Brit. J. Nutr. 19:41. Lerner, I. M., 1958. The genetic basis of selection. Wiley, New York. Lerner, I. M. and C. A. Gunns, 1952. Egg size and reproductive fitness. Poultry Sci. 31:537. 122 Liener, I. E., 1951. The intraperitoneal toxicity of concentrates of the soybean trypsin inhibitor. J. Biol. Chem. 193:183. Liener, I. E., and M. J. Pallansch, 1952. Purification of toxic sub- stance from defatted soybean flour. J. Biol. Chem. 197:29. Linerode, A., P. E. Waibel, and B. S. Pomeroy, 1961. Effects of anti- biotics, amino acids, zinc and enzymes on the growth depression in the turkey produced by raw soybeans. J. Nutr., 75:427. Lis, H., C. Fridman, N. Sharon and E. Katchalski, 1966. Multiple hemag- . glutinins in soybean. Arch. Biochem. Biophys. 117:301. T“ Lowry, D. C., I. M. Lerner, and L. W. Taylor, 1956. Intra-flock genetic merit under floor and cage management. Poultry Sci. 35:1034. Lush, J. L., 1945. Animal breeding plans, 3rd Ed. Iowa State College, Ames, Iowa. i." . Lush, J. L., 1948. Genetics of population. Mimeo, Iowa State College, Ames , Iowa. Lyman, R. L. and S. Lepkovsky, 1957. The effect of raw soybean meal and trypsin inhibitor diets on pancreatic enzyme secretion in the rat. J. Nutr. 62:269. Magee, W. T., 1965. Estimative response to selection. J. An. Sci. 24:242. Magee, W. T., 1969. Personal communication. Marks, H. L. and P. D. Lepore, 1968 a. Growth rate inheritance in Japanese quail. I. The establishment of environmental condi- tions which restrict juvenile growth rate. Poultry Sci. 47:556. Marks, H. L. and P. D. Lepore, 1968b. Growth rate inheritance in Japanese quail. II. Early responses to selection under dif— ferent nutritional environments. Poultry Sci. 47:1540. Martin, G. A., E. W. Glazener and W. L. Blow (1953). Efficiency of selection for broiler growth at various ages. Poultry Sci. 32:716. Mather, K. and R. M. Jones, 1958. Interaction of genotype and environ- ment in continuous variation. I. Description. Biometrics 14:343. ' McCartney, M. C., 1955. Heritability and genetic correlation of body weight of white Holland turkeys. Poultry Sci. 34:617. Mickelsen, 0. and M. C. Yang, 1966. Naturally occurring toxicants in foods. Fed. Proc. 25:104. 123 Newcomer, E. H., 1957. The mitotic chromosomes of domestic fowl. J. Hered. 48:227. Nitzan, Z. 1965. The effect of heating soybean meal on the apparent digestibility and metabolism of protein, methionine and lysine by cockerels. Poultry Sci. 44:1036. Nordskog, A. W., and R. E. Phillips, 1960. Reciprocal crosses involving leghorns, heavy breeds and fayoumi. Poultry Sci. 39:257. Nordskog, A. W. and M. Festing, 1962. Selection and correlated response in fowl. 12th World's Poultry Congress (Sydney, Australia, 1962) Section Papers 12:25. Osborne, T. B. and L. B. Mendel, 1917. The use of soybean as a food. J. Biol. Chem. 32:369. Rackis, J. J. and R. L. Anderson, 1964. Isolation of four soybean trypsin inhibitors by DEAE - cellulose chromatography. Biochem. Biophys. Res. Comm. 15:230. Reeve, E. C. R. and Robertson, F. W., 1953. Studies in quantitative inheritance. II. An analysis of a strain of drosophila melanogaster selected for long wings. J. Genet. 51:276. Robertson, A., 1959. The sampling variance of the genetic correlation coefficient. Biometrics 15:469. IRubel, W},D. F. Rail, and M. E. Rafter, 1967. BASTAT Program. Agri- cultural Experiment Station, Michigan State University. Saxena, H. C., L. S. Jenson, J. McGinnis, 1963. Influence of age on utilization of raw soybean rations by chickens. J. Nutr. 80:391. Schingoethe, D. J., 1969. Factors in soybeans influencing growth, pancreatic function and digestion. Ph.D. thesis, Michigan State University, East Lansing, Michigan. Sealock, R. W. and M. Laskowski, Jr., 1968. Enzymatic replacement of the arginyl residue by lysyl in the reactive site of soybean trypsin inhibitor. Fed. Proc. 27:292. (Abstract) Siegel, P. B., 1962. Selection of body weight at eight weeks of age. 1. Short term response and heritabilities. Poultry Sci. 41:954. Snetsinger, D. C. and H. M. Scott, 1958. The adequacy of soybean oil meal as a sole source of protein for chick growth. Poultry Sci. 37:1400. I'l'l 'l. I‘ll. l l. 1 '1' .‘I {1' ll 124 Thomas, C. H., W. L. Blow, C. C. Cockerham and E. W. Glazener, 1958. The heritability of body weight, gain, feed consumption and feed conversion in broilers. Poultry Sci. 37:862. Willner, D., B. Gestetner, D. Lavie, Y. Birk and A. Bondi, 1964. Soy- bean saponins. Part V. Soyasapogenol. E. J. Chem. Soc.: 5885. Wright, S., 1940. Breeding structure of population in relation to speciation. Am. Naturalist 74:232. Yamada, Y., 1962. Genotype by environment interaction and genetic cor- relation of the same trait under different environments. Jap. Jour. Genet. 37:498. Yao, T. S., 1959. Additive and dominance effect of genes in egg pro- duction and 10-week body weight of crossbred chickens. Poultry Sci. 38:284. Yao, T. S., 1961. Genetic variations in the progenies of the diallel crosses of inbred lines of chickens. Poultry Sci. 40:1048.