- . .
a..- if — I —_—_————
- '---~-_-..-9'“~ W.--_-.o-a - -- ---,,-,'__. " . - - . —————~77 .
- - -.’- - , .~ » . .- l o- -q—-—----.--vo.¢.~-op,-‘I. “..‘_‘_-‘_”‘-”'.“ IV.“'.- uvu..-.-

' ‘ “mm-'qum-ou'mahq
”-

I SIMULATION OF POTENTIAL GENETIC
‘ CHANGE IN INDONESIAN FOWL

Thesis for the Degree of M. S.
MICHIGAN STATE UNIVERSITY
MARIA ASTUTI .
1976

 

 

LIBRARY
Michigan 5‘3”
~ University

   

 

 

  
 
  

 
 
   
       

'mumue av F
"MB & SUNS'
BUOK BINDERY INC.

| [IBRARY BINDERS
;; 3:12”. . MLIICIIHI

M LA‘u‘v‘

"Hug-a" L .

 

w

 

I-..——.—

 

ABSTRACT

SIMULATION OF POTENTIAL GENETIC CHANGE
IN INDONESIAN FOWL

BY

Maria Astuti

The effect of degree of truncation selection and level of
heritability in simulated pOpulations of poultry subjected to
selection was studied.

The trait simulated was body weight at sexual maturity of
native chickens in Indonesia.

P0pu1ations were simulated for all possible combinations of
three levels of heritability (.1, .3 and .4) and three intensities of
selection by upper truncation of the distribution of phenotypes (25%,
50% and 75%).

The only genetic variance contributed to the variability was
the additive variance.

The size of offspring population in each generation was 600,
300 males and 300 females, with random mortality assumed to occur so
that breeding size after truncation was constant at 6 males and 60
females.

Selection was performed for ten generations.

Maria Astuti

For each parameter combination, 100 replications (populations)
were simulated. Two types of mating systems, random and assortative,
were simulated.

The results from random mating and assortative mating showed
the same pattern. The test of difference in slopes of regressions
of the phenotypic means on generation number for the two types of
mating was statistically significant in most cases but the difference
was biologically trivial.

The best response to selection was obtained in populations
with heritability = .4 when the proportion selected for mating was
25%. The least response to selection was for populations with
heritability = .1 when the proportion selected was 75%.

Under the same level of truncation, the response to selection

is proportional to heritability.

SIMULATION OF POTENTIAL GENETIC CHANGE

IN INDONESIAN FOWL

By

Maria Astuti

A THESIS

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

MASTER OF SCIENCE

Department of Dairy Science

1976

ACKNOWLEDGMENTS

I am deeply indebted to Dr. John L. Gill, my major professor
for his patient assistance and invaluable suggestions in the prepara-
tion of this manuscript. I would like to express my appreciation for
his guidance and encouragement in my graduate program.

I am grateful to Dr. Clinton E. Meadows and Dr. Theo H.
Coleman for their suggestions and participation in my graduate
program, to Dr. Roger R. Neitzel for his assistance in computer pro-
gramming and to Dr. William T. Magee for his sound advice and sug-
gestions.

I am also grateful for the leave of absence provided by the
Gadjah Mada University, for the financial support provided by the
Midwest Universities Consortiums for International Activities (MUCIA,
Inc.) and for the facilities provided by the Department of Dairy

Science at Michigan State University.

ii

TABLE OF CONTENTS

Page

LIST OF TABLES . . . . . . . . . . . . . . . . . . . . . . . . . iv
LIST OF FIGURES . . . . . . . . . . . . . . . . . . . . . . . . . v
INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . 1
LITERATURE REVIEW . . . . . . . . . . . . . . . . . . . . . . . . 3
Selection Experiments With Laboratory Animals . . . . . . . . 4
Selection Experiments With Simulated Populations . . . . . . 13
Various Mating Systems and Their Effects . . . . . . . . . . 22

Selection for Body Weight in Chickens . . . . . . . . . . . . 27
METHOD AND PROCEDURES OF SIMULATION . . . . . . . . . . . . . . . 29
RESULTS AND DISCUSSION . . . . . . . . . . . . . . . . . . . . . 35

Effect of Heritability on the Change of the Phenotypic

Mean . . . . . . . 35
Effect of Level of Truncation Selection on the Change
of the Phenotypic Mean . . . . . . . . . . . 45
Response to Selection with Random Mating and
Assortative Mating . . . . . . . . . . . . . . . . S4
The Genotypic Means and Variances . . . . . . . . . . . . . . 59
Estimated Accumulation of Inbreeding . . . . . . . 62
The Accuracy of the Prediction of the Phenotypic Mean . . . . 64
Estimation of Realized Heritability . . . . . . . . . . . . . 64
SUMMARY AND CONCLUSIONS . . . . . . . . . . . . . . . . . . . . . 72

LITERATURE CITED . . . . . . . . . . . . . . . . . . . . . . . . 76

iii

Table

2.1.

2.2.

LIST OF TABLES

Percent of Improvement at Generation 10 Under Random
Mating and Assortative Mating . . . . . . . . .

The Change of the Phenotypic and Genotypic Means and
Variances, and the Change of Heritability Value
Under Random Mating . . . . . . . . . . . . .

The Change of the Phenotypic and Genotypic Means and
Variances, and the Change of Heritability Value
Under Assortative Mating

Percent Estimated Inbreeding by Generation

Observed and Expected Phenotypic Means Under Random
Mating . . . . . . . . . . . . . . . . . . . . .

The Estimated Value of Realized Heritabilities

iv

Page

44

6O

61

63

65

71

LIST OF FIGURES
Figure

1. Change in the mean when b = 25% with random mating
(b = proportion selected) . . .

2. Change in the mean when b = 50% with random mating
(b = proportion selected)

3. Change in the mean when b = 75% with random mating
(b = proportion selected) . . . . . . . . .

4. Change in the mean when b = 25% with assortative
mating (b = proportion selected)

5. Change in the mean when b = 50% with assortative
mating (b = proportion selected)

6. Change in the mean when b = 75% with assortative

mating (b = proportion selected) . . . . . .

7. Change in the mean when heritability = .1 under
random mating

8. Change in the mean when heritability = .3 under
random mating

9. Change in the mean when heritability = .4 under
random mating

10. Change in the mean when heritability = .1 under
assortative mating . . .

11. Change in the mean when heritability = .3 under
assortative mating .

12. Change in the mean when heritability = .4 under

assortative mating .

13. Change in the mean under random mating and
assortative mating when b = 25%

14. Change in the mean under random mating and
assortative mating when b = 50%

Page

36

37

38

39

4O

41

46

47

48

49

50

51

55

S6

Figure

15.

16.

17.

18.

Change in the mean under random mating and
assortative mating when b = 75%

Observed and expected phenotypic mean with
random mating and b = 25% . .

Observed and expected phenotypic mean with
random mating and b = 50% . .

Observed and expected phenotypic mean with
random mating and b = 75% . . . .

vi

Page

57

67

68

69

INTRODUCTION

Many pairs of genes may be involved in the expression of
traits that have economic value. Such traits are known as quantitative
traits or characteristics involving "multiple factors," and most of
them are greatly influenced by environmental conditions.

Breeders try to improve economic traits through mating sys-
tems and selection. It is known that selection will change gene fre-
quencies, and through continuous selection over several generations,
the accumulated changes result in permanent change in the mean of the
trait selected.

The expected response and improvement made by selection can be
computed and always depends upon accuracy of selection and selection
intensity. However, different combinations of mating systems and
intensity of selection give different accuracy of selection and
selection response, and cause different results in rate of improvement.

Because substantial investment of money and time is necessary
to achieve major genetic change, considerable thinking must be done
before one makes any decision about choice of improvement program.

But quite often an improvement program will return different results
from the expected after the program has been carried out for several

years. If breeders can be given more accurate prediction of the

result of any improvement program, this problem can be overcome to
some degree.

The general method of prediction is mathematical, in which
the existing genetic theory is applied. Some modifications in the
theory may be suggested from the results of experimental studies with
laboratory animals. Relatively recently a new method that has been
called the Monte Carlo method was developed as an aid to prediction.

The random mechanisms of genetics are simulated via pseudo-
random numbers which are generated by digital computers. By incor-
porating known Mendelian laws, statistical distributions, or both,
this method can indicate what kind of result should come from dif-
ferent possible combinations of systems of mating and intensities of
selection, and establish which combination is likely to give optimum
genetic improvement.

This study involves simulation of the effect of degree of
truncation selection and level of heritability in simulated popu-
lations of poultry subjected to selection for a single quantitative
trait. The main objective was to determine:

1. the change in phenotypic and genotypic mean in each generation,
2. how response to selection in each generation will deviate

from the expectation, and
3. which selection intensity is optimal for genetic progress

at different levels of heritability in populations of a given

structure .

LITERATURE REVIEW

The genetic constitution of animals has been changing more
rapidly since they have been domesticated, as men with artificial
selection play a more important role than natural selection. It has
been shown that breeders with artificial selection and breeding sys-
tems have succeeded in changing the conformation and performance of
many species.

With the development of science and the spread of knowledge,
breeders not only seek to change animals, but also wish to predict
the amount of change and want to know how to achieve maximum change.

Breeders have always dealt with quantitative traits, but
many theories in selection to improve these traits have not been
applied. Also, there are problems in animal breeding that have not
been solved.

In order to apply the theories, to answer the questions, and
to get new results, actual selection experiments in the field must
be conducted, but time, labor, money, and facilities sometimes do not
permit controlled experiments of a size sufficient for the accuracy
required of predictions.

Laboratory animals have been used and it has been proven that
they can answer many questions about selection. Relatively recently

the Monte Carlo method of computer simulation has been used to address

questions for which suitable answers have not been obtained by other
methods. This method makes it possible to simulate genetic processes
by using repetitive sequences and pseudo-random numbers generated by
computers.

This review is mainly concerned with several selection studies
with laboratory animals and with the use of simulation procedures to
solve problems involving selection.

Selection Experiments With
Laboratory Animals

 

 

Chapman (1951) summarized the effectiveness of selection
studies involving laboratory animals. Laboratory species that have a
short generation interval can be used to answer many questions about
livestock improvement. Some results of laboratory animal experiments
have indicated that selection with or without inbreeding was effective
in producing changes in quantitative traits in both positive and
negative direction over many generations. Crosses between individuals
from positively and negatively selected lines will produce inter-
mediate offspring.

Within selected lines, the coefficient of variation usually
remained almost constant over many generations of selection. Several
experiments showed correlated response and one experiment showed a
reduction in heritability in later generations. Genetic variability
in foundation stock will be used effectively by selection and environ-
mental influence might limit the effectiveness of selection in positive

direction but not the other.

Wilson et a1. (1965) studied the pupa weight of Triboleum
castaneum as measured at 19 days after the egg was laid. Two repli-
cates were used, each replicate consisting of mass selected and
randomly selected groups. Within each group five mating systems were
applied. The selection experiment was conducted for 6 generations.

An analysis of variance of the regression coefficient of
generation mean on generation number, averaged over selection methods
and replications, revealed significant differences among selection
methods (P < .05). The test of effect of different mating systems was
conducted within the selection methods and the results showed no
significant effect of mating system among the randomly selected lines
but a slight indication of differential response among the mass
selected lines. Tests of variances showed decline over generations of
phenotypic and genotypic variation in mass selected lines regardless
of mating system. In randomly selected lines there was a tendency for
a decline in phenotypic variance but not for the genetic variance.

Another experiment with Tribolium castaneum was also reported

 

by Enfield et a1. (1966). Enfield et a1. conducted an experiment
similar to that by Rahnefeld et a1. (1963). The selection experiment

was for heavier pupa weight of Tribolium castaneum measured at let

 

day after the egg was laid. The two populations to be selected were
obtained from crosses of two highly inbred lines. The S-populations
were selected for heavier pupa weight and the C-populations were

control lines. Two replicates were used in this experiment. Thirty
six males and 72 females which were selected each generation in each

population became parents for the next generation. Individual

parents were selected on an intra-half-sib family basis. In S-
populations parents were selected from each half-sib family and in
C-populations parents closest to the family mean were selected from
each half-sib family.

There was a linear response to selection over twelve genera-
tions, but no reduction of additive genetic variance was observed.
The average unadjusted selection differential in the S-populations
was 212 ug, and after adjustment for differences in reproduction of
the individuals selected as parents, the average selection differential
was 208.5 ug.

The regressions of the difference in mean pupa weight between
S-populations and C-populations on generation time for two replicates
were 60.3 :_4.8 and 61.8 :_4.9.

McBride and Robertson (1963) carried out an experiment with

Drosophila melanogaster and combined selection for bristle numbers

 

with random mating and assortative mating. Three samples of flies
were obtained and three paired comparisons were made between random
mating and assortative mating. In two sets of lines, individual

score was used for selection and as the basis for the assortative
mating. In the third set, an index of individual and family score

was used. Selection intensity was the same in all lines. The results
showed that in all comparisons assortative mating gave a greater
selection response than random mating and the largest effect of
assortative mating was in the selected lines using index of individual

and family score. They concluded that the greater selection response

was partly due to an increase in heritability and partly due to an
increase in selection differential.
Frankham et al. (1968) also reported an experiment with

Drosophila melanogaster. In this experiment a Canberra strain was

 

used and selection was for the number of abdominal bristles. Selection
was carried for 12 generations and the objective of the experiment

was to investigate the joint effects of population size and selection
intensity. Three p0pulation sizes and five selection intensities,

with one to five replicates per combination, were used in the experi-
ment. The results indicated a considerable response and on the average
agreed well with the expectation from estimated base population
heritability, although large differences between replicates were

shown. For the same population size higher selection intensities
produced more response and for the same selection intensity the effect
of population size was not clear, but still there was a trend for the
larger population to give more response.

McArthur (1949) conducted a mass selection for small and large
body size in the house mouse. He noticed that from the first the body
size increased in mice selected for large size and diminished in
those selected for small size. After 14 generations, females two
month old of the small race averaged less than 12 grams and from the
large race 31 grams. In seven more generations the difference of 19
grams between the two races was increased to 23 grams. Males were
about 20% larger than females.

The genetics of body size in mice was studied by Chai (1956).

In his experiment he crossed large and small strains of mice to

produce F1, F2 and backcross generations. Analysis of means and
variances of 60 day body weight in the different genotypic groups was
carried out. Backcross and F2 generations had means slightly above
their respective theoretical value, F1 and F2 means were halfway
between the parents, and the backcross means were halfway between the
respective parents. Hybrid F2 had the largest variance, the second
largest was the first backcross generation, and the smallest variances
were from F1 hybrids and the large and the small strains. He reported
that the genes involved in this cross acted additively on a logarithmic
scale rather than arithmetic scale, and the dominance effect con-
tributed to the total variability, if any, was considered to be
trivial.

Miller et al. (1963) conducted an experiment to estimate the
extent of nonadditive hereditary variance in traits of mice. Five
traits were involved in the study: (1) litter size, (2) 12 week weight
of a litter of six mice, (3) individual weight at 3 weeks, (4) individ-
ual weight at 6 weeks, and (5) gain from 3 weeks to 6 weeks. Diallel
mating was used and all possible combinations of matings were made
among m males and d females. The male by female interaction component
of variance from the factorial analysis was used to estimate the
nonadditive hereditary variance. The results from 2879 mice showed
no evidence of any nonadditive hereditary variance with respect to
three and six weeks body weights. Data on litter size, obtained from

1161 litters, showed evidence of nonadditive hereditary variance,

estimated to be 28% of the total variance.

Dalton and Bywater (1963) conducted an experiment with mice.
In their experiment generation zero included offspring produced after
subsequent mating of the stocks that had been produced by four-way
crosses. Mice were placed on two different diets, "control" and
"diluted" diets. The first selection was made between the young of
generation 2 which became parents of generation 3. At weaning they
selected the whole litter of the litters that were high in either
litter size or weight. Random mating was practiced, but matings
between litter mates were avoided. No selection was practiced in the
control group. Only first parity data were collected and they observed
that litter size and litter weight were not affected by selection
over 14 generations for mice on either diet.

In the next experiment Dalton (1967) investigated effect of
selection for growth in mice on two diets, full diet and diluted diet,
where growth was measured as live weight from weaning to subsequent
mating. Selection was continued for 13 generations per year with no
overlapping generations. In each diet group mice were (1) selected
for high growth, (2) selected for low growth, and (3) randomly
selected for control. The response was expressed as deviations from
controls. All lines within each diet, except for the line selected fer
high growth on diluted diet were significantly different from control
diet.

Rahnefeld et al. (1963) performed an experiment with two
populations of mice, the S lines from the reciprocal crosses of two
inbred lines and the A line from standard laboratory inbreds. No

overlapping generations occurred in the S line but in the A line

10

generations were overlapped. Rahnefeld observed that mass selection
increased postweaning growth measured at 18 to 42 days of age.
Selection was continued through 17 generations. Mean postweaning
growth was increased 4.9 grams and the average total selection differ-
ential in the S population for males and females was 26.21 grams.

He reported that mean postweaning gain was increased about six times
the additive genetic standard deviation and about 43% of the original
mean growth.

Sutherland et al. (1968) evaluated the effectiveness of
selection in mice when combined with assortative or disassortative
mating in comparison with random mating. The characteristic measured
was body weight at 6 weeks of age. Three mating systems and three
directions of selection were applied in this experiment. The results
indicated that the mating systems had essentially no effect on the
progress from selection in either direction. This result was similar

to the result obtained with DrOSOphila melanogaster by McBride and

 

Robertson (1963) and with Tribolium castaneum by Wilson et al. (1965).

 

Furthermore, Sutherland drew a conclusion that assortative mating in
livestock will give little benefit compared to random mating, as very
few economic characters in livestoCk appear to be highly heritable.
Hanrahan et a1. (1973) investigated the joint effects of
pOpulation size and selection intensity for postweaning gain in mice
after 14 generations of selection within full-sib families. The
mating populations contained 1, 2, 4, 8, or 16 pairs. The selection
intensities were 100% (C), 50% (I), or 25% (M) saved for breeding.

The experiment was replicated. Within each population size the

11

response per generation was significantly (P < .01) larger at the 25%
selection intensity than at 50% selection intensity. Little differ-
ence in response was found between M8 and M4 and between 18 and I4.

The genetic gains in I and M16 lines were significantly (P < .01)

16
larger than in the smaller populations at corresponding selection
intensity. The larger response could be attributed only in small part
to greater cumulative selection differentials. The results suggest
that in populations of relatively small effective size, the effects of
genetic drift may cause a reduction of selection response.

Eisen (1975) examined the effects of population size and
selection intensity on long-term selection response for postweaning
gain in mice. A replicated experiment was conducted using 4, 8, or
16 pairs of parents, and selection intensities were 100% (C), 50% (I),
or 25% (M). The effects of selection for 14 generations were reported
previously by Hanrahan et al. (1973). In generation 14 the reciprocal
crosses were made between populations of 16 pairs of parents to
determine if realized heritability would be greater when selection was
applied to crosses of selected lines than in the selected lines
themselves. After one generation, each line was selected at 25%. To
observe the effect of genetic drift and inbreeding depression at the

termination of selection experiment, replicates of M M8, and M1

4’ 6
lines were crossed with C16 line. The joint effects of population
size and selection intensity on long-term response to selection for
postweaning gain in mice were similar to the results obtained for

short-term selection. Within population sizes, more intense selection

pressure led to greater total response. Within each level of

12

selection intensity, total response increased as effective number
increased. Discrepancy was observed between observation and theory.
In that experiment the total selection response did not increase
linearly as the product of effective p0pulation size and selection
intensity increased. Response to selection diminished markedly as
generation number progressed, for lines with both low (I) and high
(M) selection intensity, tending toward a plateau.

Kyle (1953) conducted an experiment for the purpose of com-
paring the theoretical effectiveness of selection for ovarian response
to a gonadotrophic hormone. Rats were selected for low and high
response on the basis of the ovary weights of full sisters injected
with a standard dosage of gonadotrophic hormone over 14 generations in
five mating systems. To predict the average ovary weight of each
litter of offspring from the average ovary weight of sisters of each
parent, a regression equation derived from a path coefficient analysis
was used, and the estimate of source of variation in ovary weight was
obtained from the foundation stock. The results of this experiment,
over all generations and mating systems, showed observed and predicted
average ovary weight from 270 litters of offspring from parents
selected for low response were 61.0 and 58.6 mg, respectively. The
difference was significant, but the observed and predicted average
ovary weights of 314 litters of offspring from parents selected for
high response were 75.4 and 76.4 mg, respectively, and were not
significantly different.

The papers reviewed in this section showed that selection for

many quantitative traits in laboratory animals was very effective

13

either for positive direction or negative direction. Selection pres-
sure and p0pulation size affected the response to selection whether
it was long term or short term selection but there were only trivial
differences in response to different mating systems.

Selection Experiments With
Simulated Populations

 

 

Scientists in various areas of research have taken advantage
of the development of high speed electronic digital computers to get
solutions for complex numerical problems at a relatively low cost.

The use of computers as sophisticated and economical tools for
geneticists became wider since Fraser (1957a) introduced simulation

of genetic systems. Fraser showed how a computer can be used to carry
out the gene by gene simulation by using binary arithmetic. Further-
more, Fraser (1957b) used the computer to simulate the effect of
linkage on rates of advances under selection. The result showed that
linkage in large populations caused a correlation of rate of advance
with degree of linkage. The effect can be neglected if the recom-
bination is greater than .5 percent. However, in small populations
the effect of linkage is considerably exaggerated.

Scheinberg (1968) divided the use of computers in genetics
into four basic categories: (1) for the design of efficient field and
laboratory experiments, (2) for the statistical analysis of data,

(3) for the numerical solutions to mathematically formulated problems,
and (4) for the simulation of real and model biological systems using
available statistical techniques. Scheinberg reported that no

information is available on the utilization of computers for the

14

construction of efficient optimum designs for genetics experiments.
The aid of computers eliminates most difficulties in analyzing an
experiment with a vast amount of data and complex computations.
Obtaining empirical solutions of complicated formulas or equations

by the use of computers has long been demonstrated. In the last
decade the utilization of computers for simulation of genetic systems
has been advanced. This simulation technique has been termed the
Monte Carlo method and many results have been reported.

Kemp and Magee (1966) used the Monte Carlo technique to
simulate selection for production traits in two breeds of swine.

The first breed was selected for feed efficiency and daily gain, the
second breed was selected for backfat probe. Crossing was made before
the first and after five generations of selection. Each trait was
affected by genes at 20 loci. The genetic correlations of daily gain
with feed efficiency and backfat probe were set at 0.6 and 0.2
respectively. Heritability for daily gain and feed efficiency was
0.3, for litter size 0.1, and for backfat probe 0.5. The results
showed that each breed improved in the traits selected. Improvements
were observed in the cross, for daily gain 0.73 to 0.75 kg/day, feed
efficiency 0.32 to 0.33 kg gained/kg of feed, litter size 8.0 to 8.2
pigs per litter, and backfat probe 3.99 to 3.66 cm.

Bereskin et al. (1969) have also conducted simulation studies
with swine parameters. Three different sizes of breeding groups were
maintained in each generation. Group 1 had 16 replicates with l sire
and 10 dams, group 2 had 8 replicates with 2 sires and 20 dams, and

group 3 had 4 replicates with 4 sires and 40 dams. The traits

15

selected were litter size and growth rate with heritability of .12
and .34 respectively. Each individual had 64 independent loci, 48
affecting both traits, eight loci affecting only the first trait, and
eight loci affecting only the second trait. Mean genotypic value for
litter size declined substantially in group 1, less in group 2 and
very little in group 3 over 10 generations. For growth rate the
genotypic mean was maintained in group 1, but in each generation of
groups 2 and 3 there were increases, averaging .25% and .50%,
respectively. There was no indication of reaching a plateau by
generation 10. Inbreeding increased at drastic rates for group 1
with the coefficient of inbreeding .76 at generation 11. Inbreeding
for group 2 and group 3 were .57 and .40, respectively. Bereskin
concluded that in small populations selection is ineffective against
random drift while in large populations random drift effect is less
and selection has more importance.

In a later paper Bereskin (1972) reported accumulated inbreed-
ing and selection effects from generation 1 to 11 for litter size in
group 1 were -10.90 i 1.48 units and 4.63 i 5.78 units, in group 2
were -7.24 i 1.78 units and 3.77 i 2.62 units, and in group 3 were
-4.33 i 0.73 units and 3.11 i 3.59 units. For growth rate he found
accumulated effects of inbreeding were -9.95 i 1.28 units, -5.96 i
1.04 units, -2.04 i 0.10 units for group 1, 2, and 3 respectively, and
accumulated selection effects were 10.68 i 6.41 units, 13.30 i 3.54
units, 15.49 i 2.33 units for group 1, 2, and 3 respectively.

Bereskin showed how small population size influenced the effectiveness

of selection. The net result after 10 generations indicated a slight

16

gain in mean genotype for growth rate in group 1, and average total
gains of 2.6 and 4.8% were obtained in group 2 and group 3 respectively.
He concluded that for traits with heritability between 0.15 and 0.35,
substantial long term genetic gains are possible even with rates of
inbreeding as high as in group 2, where a 6% loss of heterogosity
occurred per generation. The average genetic gain of individual
replicates in group 2 was 5% per generation. This response to selec-
tion was attributed to the continued availability of adequate additive
genetic variance, despite inbreeding.

Parker (1966) did a Monte Carlo simulation to examine selection
response and genetic correlation through 30 generations of simulated
selection. The trait selected for was expressed in both sexes and
each was controlled by 48 loci with equal effect at every locus. The
gene frequency was 0.5. The size of the populations of parents was
constant, 48 in each generation. The design of his experiment was
34 factorial, allowing all possible combinations of 3 degrees of
genetic correlation, 3 levels of selection and 3 levels of environ-
mental variation for either trait. Parker considered two models of
gene action, the additive model and a complete dominance model. In
the first model he found that truncation selection caused a decrease
in the genetic correlation, but it was more dependent on the level of
heritability of the selected trait than on the degree of truncation
selection if the latter was not intense. The correlated response was
directly pr0portional to that in the selected trait and depended on
the genetic covariance between the traits. In the complete dominance

model, Parker reported that when selection was by upper truncation

17

the change in the genetic correlation essentially followed the same
pattern as for the additive model.

Gill (1965a, b, c) presented a series of three papers on
simulation studies. In the first paper, he reported that in a
restricted size of population, nonrandom mating will occur and cause
an accumulation of inbreeding effect although inbreeding is not
intentional. He discussed the effect of population size upon the
results of selection under nine different models of gene action, three
of them having epistasis, conditional upon gene frequencies. The
characteristic measured was expressed in both sexes and was determined
by the genes at 40 loci equally spaced over eight chromosomes, two
alleles per loci with equal genetic effect at all loci. The effective
population sizes were 8, 12, 16 and 32 parents and selection inten-
sities were specified as 1/2, 1/4, 1/6, or 1/8. The environmental
variation was specified at 0, 1/3033, egg or 3oz,G and linkage was
specified for recombination probabilities of 0.005, 0.05, 0.2, or 0.5.

The populations with nonoverlapping generations were mating
at random. Gill assumed no effects of mutation or natural selection.
After 20 generations the mean genetic progress with complete dominance
showed that populations of 32 had the greatest mean, all pair differ-
ences were statistically significant except for populations of 16
versus 12. Through 30 generations with overdominance, comparison of
the means of all pairs of population of different size showed
statistically high significance. Under epistatic models, except
conditional epistatic, the rate of fixation was rapid and relatively

invariant with population size from 8 to 16 parents. When selection

18

intensity and heritability were high the effect of inbreeding was
larger than the amount calculated from population size. The additive-
by-dominance conditional epistatic model showed a situation where
inbreeding uplifted the genetic mean instead of depressing it.

In the next of his series of papers, Gill (1965b) reported
the effects of selection on genetic progress of finite populations.

He pointed out the evidence of ineffectiveness of prediction where
population size is finite, whether the predictions are linear or
assymptotic to the selection goal. Further he reported that both
random genetic drift and selection have great influence in changing
the values of genetic parameters.

Furthermore, Gill (1965c) reported that due to small population
size random drift hindered the progress made by selection in populations
with complete dominance, complementary factors or duplicate factors.
Selection was effective in advancing the genetic mean of small popu-
lations under all models of gene action in which the genotype of
highest merit was homozygous. However, selection was ineffective for
a character that involves only heterozygous genotypes as Optimum in
small population under mass selection. Difference in the amounts of
environmental variation between populations was important in affect-
ing the total response only when the mean was changing rapidly because
of intense selection. After the first few generations genetic merit,
gene frequency, and fixation essentially were unaffected by different
levels of linkage.

In other studies by Gill and Clemmer (1966) the effects of

selection and linkage on degree of inbreeding were simulated. They

19

created bias in estimates of the degree of inbreeding in population
of a given size. The joint effects of tight linkage and selection
can increase the rate of inbreeding severely in just three or four
generations. Selection in the absence of linkage appears to have
little effect on the variability of degree of inbreeding.

Young (1966) has also presented the results from his simula-
tion studies on the rate of decay of the additive genetic variance
due to selection, and the resulting change in heritability. He
simulated 1000 offspring each generation. The trait being selected
was controlled by ten loci, with two alleles per locus with the
initial gene frequency at each locus was .5. The design of the
experiment was a 33 factorial design and the three factors were
selection intensity, heritability value and tightness of linkage.

The selection intensities were 80%, 50% and 10% of the offspring
population saved for mating. Selection was by upper truncation. The
heritability values were .9, .4 and .l. The papulation was in linkage
equilibrium and the recombination values were .5, .2 and .05 between
adjacent loci. Selection was for thirty generations for each of seven
genetic models for all parameter sets, making a total of 189 popu-
lations.

In this paper, the first of a series, he presented only the
analyses of the additive and complete dominance models. In the
additive model the predicted and realized responses were in closer
agreement in the early generations when the initial heritability was
high than when it was low. The discrepancies between the realized

and predicted responses began to occur later in the populations with

20

high heritability but became larger than the discrepancy in the popu-
lations with low heritability after six generations. Linkage had no
apparent effect on prediction. When expected response was calculated
on the basis of the preceding generation heritability the agreement
between the realized and predicted response was close in all cases, but
the prediction was less accurate for the combination of high selection
intensity and low heritability. The rate of decay of the additive
genetic variance was most rapid when selection intensity was high and
the heritability was high. Under the dominance model predictions of
genetic advance were less accurate. Under both models, high selection
intensity with high heritability tended to overestimate genetic
advance. 0n the other hand with low selection intensity and low
heritability genetic advance was underestimated. Further, there was
no fixation of undesirable alleles under high selection intensity,

and linkage was found to have no appreciable effect on genetic
advance.

Nishida and Abe (1974) used simulation to investigate the
relationship between the skewness in the distributions of genotypic
and environmental value and the linearity of heritability in terms
of the regression of genotypic values on phenotypic values. Two
different approaches were made: (1) a deterministic approach to study
the geometric shape of heritability, and (2) computer simulation. The
results obtained from computer simulation agreed well with the results
obtained from the deterministic approach. They showed that curvi-
linearity in heritability is caused by the skewness of distribution of

genotypic and environmental values. When the distributions of

1:.-

 

21

genotypic and environmental values were not symmetric, the "linear
heritability" (ratio of ocz/opz) with the bigger value (i.e., .7)
indicated a greater deviation more from linearity compared to the
"linear heritability" with the smaller value (i.e., .5 or .3). When
selection has been practiced for many years in a closed population
and distribution of environmental values is symmetric but the dis-
tribution of genotypic values is expected to have a negative asym—
metry, the genetic gain predicted by the linear heritability under
high intensity of selection for the upper tail will be an overestimate
as the heritability will be concave downward. Further he suggested
that a practical way to use the curvilinear heritability is to divide
the data into several groups and fit a linear heritability in each
group separately.

Up to this point, results from various simulation studies have
been reviewed. The results reported by Gill (1965a, b, c) and Bereskin
(1969, 1972) were similar to the results obtained from laboratory
animals reported by Frankham et a1. (1968), Hanrahan et a1. (1973) and
Eisen (1975).

Gill (1965a) and Young (1966) have shown that simulation
studies offer the possibility to study the effect of combination of
various parameters under different models of gene action.

The above review leads to the conclusion that further develop-
ment and application of Monte Carlo methods in quantitative genetics
studies becomes more and more important as animal breeders attempt to
study complex interactions or to predict the potential response to

breeding systems before applying them to real populations.

22

Various Mating Systems and
Their Effects

 

 

Many kinds of mating systems have been developed in the fields
of plants and animal breeding. Random mating, phenotypic assortative
mating and varied forms of inbreeding and outbreeding have been used.

Li (1955) and Falconer (1960) have given formal definitions
of random mating. In the case of bisexual organisms, any one individ-
ual of one sex is equally likely to mate with any individual of the
opposite sex. In other words, the frequency of a certain type of
mating is dictated by chance.

Generally the Hardy-Weinberg Law is mentioned whenever random
mating is discussed. The law points out that under the system of
random mating a large population will be in equilibrium if mutation,
migration, and selection do not occur. Equilibrium means that there
is no change in gene frequency and genotypic proportions from
generation to generation. Also, in random mating populations the
frequencies of the genotypes in the progeny produced by such matings
are determined solely by the gene frequencies among the parents.

Phenotypic assortative mating is based on the somatic
resemblance or phenotypic likeness of parents and can be either
positive or negative. Varied forms of inbreeding and outbreeding are
the result of mating individuals because of their consanguinity
(relationship) or nonconsanguinity. Mating systems based on con-
sanguinity and phenotypic likeness differ in several ways. They will
be discussed by reviewing ground work laid by Wright and Lush.

In assortative mating the phenotypic likeness may not be genic.

If no linkage exists, all gene pairs act independently of each other

23

in consanguine mating and the number of loci does not affect the
results, whereas in somatic assortative mating gene by gene similarity
is dependent on the number of loci. Furthermore, somatic assortative
mating produces correlations between nonallelic genes which affect
the same trait, thus affecting the variability of the population,
whereas consanguine mating does not. In consanguine mating the
genetic structure of the population is not affected by heritability,
but that is not the case with assortative mating. Some experimental
studies showed that consanguine mating will cause a degree of
inbreeding that can be accumulated and may cause a depression on per-
formance.

Small population size brought about chance consanguine matings
and increased the coefficient of inbreeding in simulation studies done
by Bereskin et al. (1970) and by Gill and Clemmer (1966).

Several studies on inbreeding in swine have been reported.
Dickerson (1949) observed a linear decline in performance with
increased inbreeding for different strains of inbred swine. Bereskin
et a1. (1968) reported that as inbreeding of the litter and dam
increased, litter size~became smaller, pig mortality increased, and
growth rate was slower.

Berruecos et a1. (1970) noted that inbreeding depression was
found significant for weight of pigs at 130 days and for backfat
thickness.

Inbreeding with cattle was also reported by several authors.
Rollins et a1. (1949) studied the effect of inbreeding upon growth

in Jersey cattle. Inbreeding caused a decrease in height, weight and

24

heart girth and there was a linear regression on percent of inbreeding.
Records were made on 322 cattle, with average inbreeding of 15%.
Significant decrease was shown in birth weight. The effect of
inbreeding varied with age and characteristic. The maximum effect
occurred at 6 months of age where an increase of 1% inbreeding caused
a decrease of 0.4% in mean weight, 0.15% in mean heart girth, 0.16% in
mean height. At 4 1/2 years this figure changed to 0.10% in mean
weight, 0.01% in mean heart girth, and 0.01% in mean height. Further,
they noted that until six months of age inbred animals grew more
slowly than out-crossed animals but between 6 and 12 months they grew
more rapidly and continued to do so for the rest of the period studied.

Nelms and Stratton (1967) found that inbreeding in Hereford
calves caused a decrease of .465 kg for each percent increase in
coefficient of inbreeding for 180-day weight.

Dinkel et a1. (1968) have also reported that inbreeding of
calf and inbreeding of dam effects are of more importance on weaning
traits than on postweaning traits and the two sexes of calf showed
different response. Further, they reported that growth was more
affected than conformation by inbreeding.

The effect of inbreeding on production in Holstein cows was
examined by Von Krosigk and Lush (1958). For each increase of 1%
inbreeding intra-sire regressions were -l.74 i .57 1b. of butterfat,
~54 i 17 lb. of milk, and +0.003 i .003% of butterfat, when the
average of each cow's records was used.

Shoffner (1948) studied the reaction of the fowl to inbreeding.

Using 9 lines he found that the general effect of inbreeding was

25

depression, but that the performance trait depressed most differed
from line to line. The greatest reduction in performance has been

in hatchability, egg production, viability, and sexual maturity.

There were no changes in body weight nor egg weight. With inbreeding
up to 60%, the regression of hatchability of fertile eggs on coef-
ficient of inbreeding (Fx) was -.436 i .132. Egg production was
measured as the number of eggs laid by an individual until she is 500
days of age. The within sire regression of the mean egg production of
sister group of mean inbreeding was -.926 i .068. The regression of
sexual maturity (days from hatching to first egg) on Fx was .597 i .111,
the regression of body weight (pounds at 300 days of age) on Fx was
-.004 i .003, the regression of average weight of ten eggs was -.002 i
.008.

MacLaury and Nordskog (1956) studied the effect of inbreeding
on mortality in the domestic fowl. Regression of mortality percent on
inbreeding was 0.33 in the brooding period, 0.15 in the range period
and 0.21 in the laying-house period. Some of the disadvantageous
effects of inbreeding were shown. In general the effect of inbreeding
was the reduction of phenotypic mean value of some characters.

Lush (1948) and Falconer (1960) pointed out that one can
employ the effect of inbreeding as a tool to obtain heterosis. No
reduction in chicken body weight due to inbreeding was reported by
Schoffner (1948), and other reports by Martin et al. (1953), Brunson
et a1. (1956) and Kan et a1. (1959), showed that body weight was an
additive trait. In this case there was agreement as inbreeding

depression should not occur with additive traits.

26

On the other hand if dominance involved in body weight, as
stated by Siegel (1962) and Yao (1959), inbreeding depression will
occur and mating two inbred-lines offers a possibility to obtain
heterosis.

Somatic assortative mating is as Falconer (1960) describes it,
when mated pairs tend to be of the same genotype more often than could
occur by chance. Normally the mating so described is called positive
assortative mating. Less often negative assortative mating is used,
in which the tendency is toward the mating of genetic Opposites.

In Lush‘s (1948) review, he stated that mating like phenotypes
tends to bring together mates whose genes have similar effects but
need not be allelic to each other. The likeness between the pheno-
types can be high while yet the likeness between mates gene by gene
can be very low. As the number of gene pairs that control a trait
increases, less and less of phenotypic correlation comes from likeness
in allelic genes and more and more of it comes from the nonallelic
genes. Positive assortative mating tends to throw the population
toward the two extremes and to diminish the frequency of the inter-
mediate types, and negative assortative mating will show the opposite
results. Moreover assortative mating can be practiced only for
characteristics which can be seen and measured. It is effective in
increasing variance for certain characteristics only if they are highly
heritable and controlled by few genes. Some experiments with assorta-
tive mating have been done with laboratory animals.

Wilson et al. (1965) studied Tribolium, Sutherland et a1.

 

(1968) mice, and McBride and Robertson (1963) Drosophila melanogaster.

27

The review of these studies has been made in the previous section
with the conclusion that assortative mating gives some advantage when

the trait selected for is highly heritable.

Selection for Body Weight in Chickens

 

In chickens for broiler production, body weight is the most
important trait, and profit will be more if breeders can improve this
trait through selection.

Various results from selection for body weight have been
reported. Godfrey and Goodman (1955) reported that selection,
within line, for small and large body size in Silver Oklabar after
5 generations showed differences in six and 12 week body weights of
0.6 and 1.3 pounds, respectively. Heritability average was about
0.26 at both 6 and 12 weeks of age. Phenotypic variation was 14%
for the large line and 19% for the small line, but genic variance of
the large line was greater than that of the small line.

Siegel (1962) reviewed heritability studies for body weight
in chickens. He noted that 176 published heritability estimates of
body weight obtained for chickens 6 to 12 weeks of age gave the
quartile range of .29 to .54 with the median of .41. Siegel (1962)
conducted a selection experiment for body weight at 8 weeks of age in
White Plymouth Rock chickens. He reported that 30% of the total
variation was due to additive genetic variation, and epistatic devi-
ations appeared to be unimportant. Siegel found that dominance and
maternal effects influence body weight, as was reported by Yao
(1961). There was an indication of sex-linked gene effects as

reported by Brunson et a1. (1956) and by Thomas et a1. (1958).

28

Martin et a1. (1953) reported that variance in body weight
was due to additive genetic variance and random environmental factors,
and that dominance effects did not show statistical significance.

Brunson et a1. (1956) found that variances in body weight due
to genetic differences were 41% for additive genes, 2% for nonadditive
genes, and 10% for sex-linked genes. The effect of sex-linked genes
may be considered important as it was also reported by Thomas et a1.
(1958).

Yao (1959) reported that dominance effects were found to be
significant in several incrosses and in crossbreds. In a later study
Yao (1961) found that maternal effects had an influence on 10 week
body weight.

Kan et a1. (1959) performed selection studies on nonadditive
gene effects on six broiler traits. Results showed trivial, if any,
nonadditive gene effects on body weight at either 4 or 9 weeks of age
or on gain in weight from 4 to 9 weeks, but nonadditive effects were
important contributors to variation in shank length, keel length, and
body depth. Heritability for body weight was quite high (Siegel,
1962), and the variance of body weight was mainly due to additive
genetic variances. Further, Martin et a1. (1953) indicated non-
additive genetic variance has little or no effect. The result was
confirmed by Brunson et a1. (1956), and by Kan et a1. (1959), so
mass selection should be successful to bring an improvement in body

weight.

METHODS AND PROCEDURES OF SIMULATION

Computer processes were used in this study to simulate the
base parental population. A repetitive pseudo-random number generator
was utilized in simulating the probabilistic genetic mechanisms.

Two mating systems were used, random mating and positive
phenotypic assortative mating. Individuals in generation one in both
systems were generated from random mating of the base parental popu-
lation.

To obtain the data for this study FORTRAN programs with
several subroutines were written and used in a CDC 6500 computer, at
Michigan State University.

The programs in detail were too lengthy to be listed. Only a
brief explanation of how the populations were simulated will be
given.

The population simulated in this study was the native chicken
population from the author's country (Indonesia) and selection was
only for body weight at sexual maturity. Body weight of the native
fowl is easy to measure, is likely to be moderately heritable, and is
economically important in Indonesia.

The native chickens are raised widely but not intensively in
the villages in Indonesia and their characteristics have not been

studied thoroughly. However their potentiality in contributing to

29

30

the animal protein resource is great. Up to now no purposeful selection
has been practiced and no research information has been obtained.

This simulation process is based on reviews of literature con-
cerned with other populations of fowl. It has been assumed that the
variance of body weight at sexual maturity is only due to additive
genetic variance and environmental variance, i.e., that dominance and
epistasis play no more than a trivial role in the determination of
variation in body weight.

The design of this experiment is 32 factorial. Populations
were simulated for all possible combinations of three levels of
heritability and three intensities of selection by upper truncation
of the distribution of phenotypes.

In each base population the parameters used for mean and
standard deviation of body weight were 1.4 kg and .2 kg, to closely
approximate the real conditions in Indonesian villages.

The three levels of heritability selected were .1, .3 and .4,
and as the additive variance was the only genetic variance contributed,
the corresponding value of environmental variance can be calculated
easily.

The three levels of selection by truncation were 25%, 50%
and 75% of young birds saved for breeding. For each pOpulation,
selection was practiced for ten generations and 100 replication popu-
lations were simulated for each combination of heritability and

selection intensity.

31

The FORTRAN program was made general for all populations by
permitting variable "input" in each run for the heritability value
and level of selection.

The size of offspring population in each generation was 600,
300 males and 300 females.

In the process of growing until birds reached sexual maturity,
mortality was assumed to be random with respect to genetic potential
for body weight. Random mortality here was defined as the reduction
of the population size in the real condition because of weakness,
diseases, predators, slaughter or sale.

After the appropriate level of selection by truncation was
applied the numbers of males and females available for mating was
always 6 and 60, respectively.

The process of simulation can be summarized as follows:

(1) Each run of the program started by reading the herita-
bility value and the level of selection.

(2) A library program, RANF, was available at MSU for the
generation of uniformly distributed random numbers.

In this study uniformly distributed random numbers in the
range 0‘f_ri_§ +1 were generated by the RANF program.

The variance of a uniformly distributed variable is equal
to the square of the range divided by 12.

Therefore, V(ri) = (1)2/12 = 1/12

Twelve random uniform numbers (ri) were added together to

form a random sum in the range 0 to 12. Then, the variance of the

32

sum was equal to 12 L%§l-= 1, so the standard deviation of the sum
also was 1.

Upon subtracting six from any sum of twelve ri, one obtains
random numbers in the range, -6 f_ej'f_+6, and the ej are normally
distributed with a mean of zero and standard deviation of 1, providing
standard normal random deviates.

A sample of random deviates so generated was examined and
found to conform closely to the standard normal distribution.

(3) A subroutine that calculated the genetic and environmental
deviation was called in the main program.

Six males and 60 females were simulated as the base popu-
lation.

The genotypic and phenotypic values were calculated as
follows: genotypic value = pOpulation mean + genetic deviation;
phenotypic value = genotypic value + environmental deviation, where
genetic deviation was calculated as (Std. Dev. of additive genetic
values) (Random normal deviate) and environmental deviation was cal-
culated as (Std. Dev. of environmental effects) (Random normal
deviate).

(4) The second subroutine, which generated the offspring
from randomly mating each male to ten females was called next. From
those matings the first generation for selection was obtained.

Random matings produced 5 males and 5 females each, to insure
that the numbers of males and females were equal. The total offspring

in the first generation consisted of 300 males and 300 females.

33

The genotypic values of offspring were calculated as:

Genetic value of sire + Genetic value of dam
2

 

+ random deviate from Mendelian sampling, and the phenotypic values
were computed from genotypic value of offspring + environmental
deviation, where the Mendelian deviate was calculated as V75 x (Std.
Dev. of additive genetic) x (Random normal deviate).

Next, a subroutine was called to simulate random mortality,
which was assumed to occur before the offspring reached sexual
maturity.

The numbers of male and female individuals that survived were
determined by the desired level of truncation selection, so that after
the selection was practiced the numbers of offspring saved and retained
to be parents for the next generation were always 6 males and 60
females.

In this process a random number was generated and multiplied
by 300, producing a number in the range of 0 to 299.999, which by the
addition of 1 gave a number in the range of l to 300.999. This
number then was truncated to integer value resulting in a random
integer in the range 1 to 300. The same process was followed for
the random survival of female individuals.

(5) The last subroutine was called for sorting individuals
in descending order of phenotypic value, so that they were ready for
selection. For both random mating and assortative mating the

simulation process was the same until this point in the program.

34

In assortative mating the t0p male was mated with the ten
top females, the next top male to the next of the ten t0p females,
etc. But in random mating another subroutine put the individuals
for mating back in random order.

(6) To obtain ten generations of offspring a "do loop" was
made in the main program following generation of the base population.
Another "do loop" that was built into the main program, after the
process of reading heritability value and the level of selection,
made it possible to obtain 100 replications (simulated populations).

(7) Results were stored on magnetic tape to avoid excessive
printing.

A second program was made for the statistical analysis of
data stored on tape. The results printed from this program were the
phenotypic and genotypic means and variances of 100 replications
(populations) in each of 10 generations. Computing time for this
simulation program was approximately one hour for each set of 100

populations.

RESULTS AND DISCUSSION

For each type of mating system (random and assortative), nine
combinations of heritability and level of truncation selection were
used. For each combination 100 populations (replicate runs) were
simulated. '

In each run the base population was generated at random using
a repetitive pseudo-random number generator. Each population was per-
mitted to pass through ten generations of selection and mating.

The final results that were printed out were the phenotypic
and genotypic means and variances across 100 replicates (populations)
in each generation.

Effect of Heritability on the Change
of the Phenotypic Mean

 

 

The change in the phenotypic mean under random mating is shown
in Figures 1, 2 and 3, where phenotypic mean is plotted against the
generation. The change in the phenotypic mean under assortative mating
is shown in Figures 4, 5 and 6.

The results when the proportion selected for breeding was 25%
(Figure l for random mating and Figure 4 for assortative mating),
showed that the highest heritability, .4, gave the greatest change in

the phenotypic mean from one generation to the next and gave the

35

Means (kg)

 

1.3

 

36

 

I-O-o-I—n 0—

‘ AL A

1 2 3 4 5 6 7 8 9 10

GENERATION

Fig. 1: Change in the mean when b=25Z with random mating
(b - proportion selected)

h2=.1
h2=.3
h2=.4

Means (kg)

37

2.0 v
1.9 I
.I
I
’l
1.8 I/{/
//
Jr
/ _/.
I, ./
1 7 I If /

l.

1.6 . ,’/"/,/’
I,

 

 

3

 

; A . A A I . . .

1 2 3 4 5 6 7 8 9 10
GENERATION

Fig. 2: Change in the mean when b=50Z with random mating
(b = proportion selected)

Means (kg)

2.0

1.9..

1.8

1.7

1.3

 

38

 

 

f f

1 2 3 4 5 6 7 8 9 10

GENERATION

Fig. 3: Change in the mean when b=7SZ with random mating
(b = proportion selected)

h2=.1

-.-._.__.._ h2=.3

h2=.4

Means (kg)

2.2

2.1

2.0

1.9 .

1.8

1.7

1.3

 

39

I,
I
. J
// //
, ,/ ,/'
I ./

h2-.1
h2-.3

1.-_-____.._ 112-: . 4

 

—. -o—0— Ou—

 

 

1 2 3 4 5 6 7

0
1
H

v v fl,

9 10

on

GENERATION

Fig. 4: Change in the mean when b=ZSZ with assortative mating
(b = proportion selected)

Means (kg)

2.0

1.8

1.7

1.3

 

40

112:. 1
_......._...._.._. 112:. 3

....... __ h2=.4

 

1234 5 67 8 910
GENERATION

Fig. 5: Change in the mean when bBSOZ with assortative mating
(b - proportion selected)

Means (kg)

2.0

1.9

1.8

1.7

1.3

 

41

h2=.1

 

—C_ 0—0— o- 112:. 3

__..._.__._... h2-.4

 

v V V '

113 4 5 67 8 910

GENERATION

Fig. 6: Change in the mean when b=752 with assortative mating
(b - proportion selected)

42

highest response at the end of the selection studies for either mating
system.

Figures 2 and 5 show the results when the proportion selection
was equal to 50% and Figures 3 and 6 show results when the pr0portion
selected was equal to 75%, for random and assortative mating, respec-
tively, at each level of selection.

Results under random mating and assortative mating showed the
same pattern, that the higher the heritability the higher will be the
change, i.e., the greater improvement by selection. Differences
between the two mating systems were minor.

In the two mating systems of this study, the standard error
of the phenotypic mean was obtained by the following calculations.

The computer printout gave only the phenotypic variance of all

individuals. The calculation of standard error of the mean is:

Yi = individual value (result)
Y2 = mean value Of 600 individuals in one population (replication)
= (600 )
2 Y./600
. 1
1=1
Y3 = mean value of 100 populations (60,000 individuals)
V(Yi) = variance of all individuals
= 0.04
V(Y2) = variance of mean of 600 individuals (one population)

43

600
Z (0.04/360,000)
181

600 (.04!

360,000 = .000067

V(Y

variance of the mean of 100 populations

100
z (.000067)/10,000
i=1

Standard error = [.000067 = .0008
100

The effect of large replication is shown in this result, as the value

of the standard error is very small. The value of the standard error
of the mean obtained under random or assortative mating was equally
small (. 0008) .

In both systems of mating the percent of improvement in the
phenotypic mean at generation ten was calculated as:

Phenotypic mean at generation ten - Phenotypic mean at generation 1

Phenotypic mean at generation 1 x 100‘

The values are presented in Table 1.

For 25% selection under random mating the improvements at
generation 10 were 14.3%, 39.5% and 53.6% for heritability equal .1,
.3, and .4. When 50% were selected the corresponding improvements were
9.4%, 25.4%, and 32.6%, and when 75% were selected improvements were
4.6%, 14.5%, and 18.1%.

44

Table 1.--Percent of Improvement at Generation 10 Under Random Mating
and Assortative Mating.

 

Mating Type
Random Mating Assortative Mating

 

Proportion Selected

25% 50% 75% 25% 50% 75%
h = .1 14.3 8.4 4.6 14.0 9.2 4.6
h = .3 39.5 25.4 14.5 41.3 26.9 14.4
h = .4 53.6 32.6 18.1 54.3 34.8 18.6

 

For assortative mating improvements were 14.0%, 41.3%, and
54.3% when 25% were selected, 9.2%, 26.9%, and 34.8% when 50% were
selected, and 4.6%, 14.4%, and 18.6% when 75% were selected, for
heritability equal to .l, .3, and .4 in each case.

These results showed good agreement with the theory that the
response to selection can be predicted and depends on the intensity of
selection, heritability, and phenotypic standard deviation of the
trait selected.

Falconer (1960) gave the prediction equation of response to

selection as:

where i is the intensity of selection, h2 is the heritability value
and OP is the phenotypic standard deviation. The value of i for
different proportions selected for mating can be obtained from tables

presented by Lush (1956) and by Nanson (1967).

45

Under the same selection intensity and the same phenotypic
standard deviation the response to selection will be proportional to
the heritability.

It was shown that the highest heritability (.4) gave the
highest response at the end of selection. There have been no reported
research studies on heritability of body weight of the native Indonesian
chickens. Siegel (1962) reported the median heritability estimate of
body weight for chickens 6-12 weeks of age is .41.

If the heritability of body weight in native chickens at sexual
maturity were assumed to be .4 instead of .3 or .1, then, the improve-
ment that can be achieved in 10 years can be estimated accordingly.
Under random mating with 50% of population saved for breeding in 10
years, the heritability .4 will create a population mean difference
in improvement of 106 grams above that for heritability .3. Between
the heritability .4 and heritability .l the difference will be 343
grams.

Effect of Level of Truncation

Selection on the Change of
the Phenotypic Mean

 

 

 

The effect of selection intensity on the change of phenotypic
mean in each generation for heritabilities .1, .3, and .4 is presented
in Figures 7, 8, and 9 for random mating and in Figures 10, 11, and
12 for assortative mating.

As expected the results showed that the higher the level of
truncation (the smaller the population selected for breeding), the

greater the change in each generation and the more the response at the

 

 

Means (kg)

1.

3

 

46

 

II
N
UI
o\°

 

U‘ U" 0"
II ll

\1 UI

U1 0

N N

 

 

Fig.

7:

GENERATION

Change in the mean when heritability
under random mating

Means (kg)

 

 

47

b=25%

 

_.-O__Q-_

GENERATION

Change in the mean when heritability =
under random mating

b=50Z
b=75%

.3

Means (kg)

1.4

1.

3

 

 

Fig. 9:

48

b=25Z
,_,_b=50%
b=7SZ

 

.8‘..

3 4 5 6 7 8 9 10

GENERATION

Change in the mean when heritability = .4
under random mating

Means (kg)

 

Fig. 10:

34567 3 9 10
GENERATION

49

 

Change in the mean when heritability =
under assortative mating

 

.1

Means (kg)

2.0.

1.3

 

 

50

b=25Z
._ . __._.__ b=50%
._b=75Z

 

GENERATION

Fig. 11: Change in the mean when heritability = .3
under assortative mating

Means (kg)

 

51

 

 

 

 

Fig. 12:

Y i T V

4 5 6 7 8 9 10
GENERATION

Change in the mean when heritability =
under assortative mating

.4

52

end of selection. This was true both for random mating and for
assortative mating.

These results showed good agreement with those of Frankham
et a1. (1968), Hanrahan et a1. (1973), and Eisen (1975).

For either random mating or assortative mating under the same
level of truncation, the response to selection is proportional to
heritability.

The best response to selection at the end of selection was
obtained in populations with heritability = .4 and the proportion
selected for mating was 25%. The least response to selection in this
study was for populations with heritability = .l and proportion
selected was 75%.

Each combination of heritability and proportion selected for
mating gave different response. An analysis of variance was made to
see if the level of truncation and the heritability level gave a
significant contribution to the variability of response to selection
at the termination of studies. Data for this analysis of variance
were phenotypic mean values at generation ten under random mating.

The analysis of variance table is presented here:

Source of variation df S.S. M.S. F
Heritability 2 .203378‘ .101689 l5l7.746**
Truncation level 2 .161588 .080794 1205.88l**
"9" x truncati°n 4 .397318 .0993295 l482.530**
level

Error 891 .000067

Total 899

53

The F statistics were tremendously large, the heritability,
truncation level and the interaction between them being highly signifi-
cant in contributing to the variability of response to selection. The
mean differences between each level of truncation within the same level
of heritability were tested with the Bonferroni t statistics. The
Bonferroni t statistics gave the confidence interval of the mean
difference.

C.I. = :_(3.280)(.0011575) = :_.0038 for a = .1, and the test of the
mean difference between level of truncation within each level of
heritability was significant.

The best response to selection was shown to be where 25%
population was saved for breeding. That was true for all values of
heritability applied in this study.

With random mating, after ten years selection the difference
of the phenotypic mean if 25% population is saved for breeding instead
of 50% amounts to 289 grams (assuming that the heritability a .4).
Between 50% and 75% the difference is 204 grams. The difference in
amount of progress shown is meaningful but in the practical situation
50% or 75% level of truncation seems more likely to be achievable
than 25%. Some reasons can be given.

1. The idea of truncation selection is new to the village peOple;
it needs a careful introduction with the least possible risk.
2. When truncation selection is applied the natural condition will

be changed.

54

3. People will be more likely to accept the smaller change. In
this case it will be easier to accept the idea to save 50% or
75% of the population for breeding than 25%.

4. Reinforcement of the value of selection will be slow, i.e.,
selection takes a long time before the major results can be
observed.

5. Natural selection still exists and disease control is minimal.
The risk will be lesser if the population saved is bigger, so
that populations can be continued.

6. Fertility in native chickens probably is low. The more that
can be saved, the more assurance for the continuation of
selection.

If the level of truncation were 50%, the expected progress in
10 years would be 459 grams, nearly double the progress made by 75%
truncation (255 grams).

Considering that the raw material of selection is genetic
variation, to apply 50% level of truncation seems suitable when the
population is considered large enough, but when the population is
smaller, 75% level of truncation will be preferable.

Response to Selection with Random
Matingand Assortative Mating,

 

 

The response to selection when mating was assortative was
nearly the same as when mating was random. Figures 13, 14, and 15
show phenotypic means under random mating and assortative mating,
plotted against generation numbers, for 25, 50 and 75 percent truncation

selection, respectively. The differences in slopes of regressions of

Means (kg)

[—1

p.-

55

@ Random mating
‘ Assortative mating

 

 

 

h2=.l
__,___“__,___h2=.3
h2=.4

 

V V Vi

2 3 4 5- 6 7 8 9 10
GENERATION

Fig. 13: Change in the mean under random mating
and assortative mating when b=25Z

Means (kg)

2.1

2.0

1.9

1.8

1.7

1.6

1.5

1.4

1.3

 

56

 
  

9 Random mating
Al Assortative mating

h2=.1
h2=.3
h2=.4

 

2 3 4 5 6 7 8 9 10
GENERATION

Fig. 14: Change in the mean under random mating
and assortative mating when b=50Z

Means (kg)

2.0

1.9

1.8

1.7

1.6

1.5

1.4

1.3

 

57

 
  

Q Random mating
4i Assortative mating

 

4r

A ‘
v Vi r

23 4 56 7 8 910

q
q
q

GENERATION

Fig. 15: Change in the mean under random mating

and assortative mating when b=75Z

58

the phenotypic means on generation number for the two types of mating
was statistically significant in most cases but the difference was
biologically trivial except when selection intensity and heritability
were high. In that case a small advantage may be noted for assortative
mating. In this study statistical significance has little meaning.

The difference that can be detected is very small because of large
sample size (60,000 individuals per mean).

Sutherland et a1. (1968) reported that assortative mating
gives little benefit compared to random mating if the trait is not
highly heritable. The results of this study agree with that. When
the heritability was .1, assortative mating and random mating gave
nearly the same results regardless of level of truncation selection.
When the proportion selected for mating was 75% the results from both
types of mating were very close for each of the three levels of
heritabilities. Possibly, assortative mating in small populations
tends to produce more genetic variation than in the case of random
mating, but when the population saved for breeding is relatively large
(as in the case of 75% selected), then the genetic variation may be
nearly the same for both types of mating. In general, it appears that
there is little or no benefit to assortative mating over random mating.
Furthermore, in the practical situation with which this study is con-
cerned, it would be very impractical if not impossible to conduct
assortative mating. The native chickens in Indonesian villages are
not raised in confinement, so even if assortative mating gave a

meaningful benefit, still such matings could not be adopted unless

59

tremendous effort was made to change the traditional conditions in the

villages.

The Genotypic Means and Variances

 

The genotypic means and the changes from one generation to the
next, for each of the two types of mating systems, were very close to
the results observed for the phenotypic means. The differences were
trivially small if not zero.

The closeness of these results was expected for the trait
simulated because the average environmental effect was expected to be
zero.

Both phenotypic and genetic variances increased slightly from
one generation to the next, but the genetic variance increased slightly
more than the phenotypic variance. That caused heritability to
increase somewhat over time. The heritability values increased more
with assortative mating than with random mating when selection was not
intense (75% or 50% saved for breeding), for each of the three initial
levels of heritability. For more intense selection (25% saved), the
changes in heritability were similar for the two types of mating.

The standard errors for the genotypic means were between .0003
and .0006.

Table 2.1 and 2.2 show the phenotypic and genotypic means and
variances obtained by random mating and assortative mating, respec-
tively. As in general there is little or no difference in practical
benefit between assortative mating and random mating, only random

mating will be discussed further.

60

 

 

Ame. eeeeNe. Remeee. mme.A mme.A Rm». neemAe. eemmee. eee.A mee.A eNA. «Renee. mAeeee. mee.A nee.A eA
Awe. AeeeAe. unemee. ANe.A eme.A een. emAeAe. eAeNee. nem.A Nem.A ANA. eeeeee. AAAeee. eme.A eme.A e
Ree. AAmeAe. AeAeee. Aee.A Aee.A New. AeNeAe. ReeAee. eem.A eem.A ANA. eeeeee. neeeee. mee.A eve.A .e
eNe. eNAeAe. eeNeee. mem.A eAm.A ewe. eemnAe. eeeAee. emm.A enm.A ANA. eeeeee. eAAAee. ene.A eee.A A
mee. mAaeAe. mANAee. nem.A eem.A mAn. meeNAe. NAAAee. mAm.A mAm.A eAA. eaeeee. memeee. mne.A mne.A e AAA
eee. AemeAe. memeee. mAm.A eAm.A AAA. eAeNAe. mnmAee. eee.A mee.A 4AA. Rmeeee. eeeeee. eNe.A eme.A m
eem. eNemAe. meeeee. eee.A eee.A men. eAAAAe. Ameeee. ehe.A eAe.A AAA. enmeee. eneeee. AN4.A ~N4.A 4
new. mmNmAe. mAmeAe. eme.A eme.A new. mAeNAe. meeeme. Ame.A Ame.A AeA. AAmeee. Neeeee. eAe.A eAe.A m
eem. eeAmAe. meNeAe. mme.A Nee.A AeA. AeeNAe. eeNeee. eNe.A eNe.A eeA. eeAeee. Aheene. eee.A eee.A N
eem. AmamAe. eeeeme. Aee.A Aee.A een. eeNNAe. eeneee. een.A eee.A eeA. mAeeee. Neeeee. eem.A eem.A A
eAe. emmAAe. eeeAee. eme.A eme.A eNA. ememAe. NeeAee. Ame.A mmA.A eAA. emeeee. emheee. eAm.A eAm.A eA
eAe. NeNAAe. AeNAee. Ree.A Ree.A ANA. mm~mAe. RemAee. AAA.A AAA.A eAA. eaeeee. eaeeee. nem.A Nem.A e
NAe. meeeAe. Reeeee. AeA.A Nee.A een. enmNAe. meeeee. mee.A see.A eNA. eeeeee. emeeee. Aee.A eee.A e
nee. eeAeAe. meAeee. AAA.A AAA.A eAm. eAANAe. eAeeee. ene.A Ame.A 4AA. mmeeee. meeeee. Aee.A Aee.A A
new. eemmAe. meeeee. eee.A eee.A men. meNNAe. neNeee. eem.A eem.A AAA. eeeeee. eemeee. eee.A eee.A e see
new. enemAe. AAeene. AAe.A AAe.A Aen. AmeNAe. meeeee. eem.A Aem.A AeA. AAAeee. eAeeee. eme.A eme.A m
een. eeemAe. Ameeee. eem.A eem.A Aem. eAeAAe. meaeme. eAm.A ANA.A AeA. eeeeee. eeeene. see.A eve.A e
wen. memeAe. eemene. eAm.A eAm.A eee. meeAAe. Reeeme. Nee.A Nee.A AeA. mAeeee. AAeeme. e~4.A eNe.A A
Aem. ANeeAe. emeene. eme.A emv.A heN. AeAAAe. mmeeme. mee.A nee.A NeA. eeeeee. Ameeee. eAe.A eAe.A A
nee. RmAeAe. meeeee. Nee.A Nee.A mew. eAeAAe. mNeeee. eem.A eem.A NeA. AAAeee. eAneee. eeA.A. eem.A A
ewe. emeeAe. eeeuee. eeA.~ eeA.~ Ann. eemnAe. AeeAee. nme.A nee.A AAA. Neemee. eemAee. Aee.A eee.A eA
eNe. eNNeAe. eeemee. eee.~ Ree.~ 4AA. eeeNAe. NAeAee. nee.A Nee.A eNA. memmee. eNeAee. Ahm.A enm.A e
eAe. ANAAAe. emnAee. eee.A eee.A Rem. NAANAe. NmAeee. Nee.A Ane.A eNA. Reemee. AemAee. Amm.A Amm.A e
eAe. eeeeAe. eeeeee. eee.A eee.A een. eeANAe. Aeeeee. NAA.A ~AA.A eAA. Reeeee. ANeeee. emm.A eNA.A A
Nee. eANeAe. Reneee. Ame.A Ame.A New. eeeNAe. emeene. eAA.A AAA.A 4AA. mneeee. eneeee. eem.A eem.A 8 .mm
eem. NeAmAe. Aeeeme. eeA.A eeA.A New. eeeAAe. ANNeme. eme.A mme.A eeA. eemeee. AAAeee. eee.A eee.A m A
new. eAemAe. meeeee. mee.A mee.A AeA. AmmAAe. eneene. mem.A eem.A meA. eAeeee. memene. eee.A nee.A e
Aee. NeAeAe. eeeeme. nem.A enm.A Nee. eeeeAe. eNeeme. Anm.A emm.A neA. AAAeee. Ameeme. nee.A eee.A A
New. eNAeAe. mAmeme. mee.A eee.A AeA. eNeeAe. «Neeme. nee.A eee.A AeA. Neeeee. emAeee. eNe.A m~4.A A
eee. eAeeAe. Aneeee. eem.A een.A eeN. eeeAAe. meAeme. eem.A eee.A ee.e AAemee. emAeee. eee.A eee.A A
N..— . “C00 .Cflr—Q . HCOU . GOfim NS . UGGU . COr—Q . HCOU . CO£Q N; . UCQU . COSQ . UCQU . GOA—Q COMHMHQr—OU EHUOHOW
005MHG> :dG: OUCGMHQ> :50: OUGdHHG> CQOZ COHHHOQOHQ
e. n. A. AAAAAeeeAAee .

 

.mewue: aopcmm have: osAm> Aufiawnmuwuoz mo omcmcu can use .mooemwue> pee memo: uwmxuocou use ufimxuocozm Ozu mo omcmnu och--.A.~ OAQMH

61

 

 

omv. cummmo. nwwovo. coo.A ooo.~ van. oeNAAo. swomvc. moo.A voo.A NnA. Hammoo. mmmovo. mow.“ wov.A oA

mac. mmmmwo. enmoeo. nno.A Nno.A won. nonAc. nuance. Nam.A ~wm.A emu. wmAmoo. nmoAvo. mmv.A wmv.A a

new. mNNAwo. mwwvvo. noo.A voo.~ can. ouemAc. bounce. Hom.A oom.A mNA. mwomoo. canceo. mme.A omv.A w

awe. enmomo. awnevo. onm.A Anm.A Non. cAemAc. mmenvo. mmm.A mnm.A nNA. noomoo. oomovo. mvv.A mee.A A

ewe. vmmomo. mmnveo. ovm.A oem.A new. onvAo. nnAch. nAm.A nAm.A oNA. Amwvoo. Ammovo. env.A one.A c «we
mmv. maoon. owonvc. AAm.A AAm.A own. mAvao. mmmmeo. emv.A mmv.A oAA. moveco. mmmovo. omv.A Amv.A m

wnv. anomao. mNcho. owe.u owv.~ Nnm. owwmfio. mowuvo. use. use.“ aca. noveoo. oNvovo. NNv.~ NNv.A v

mAv. mmmOAo. mweoeo. ome.A omv.A wAn. AmAnAo. omNAvo. mee.~ eve.A ecu. mnAvoo. Acoovo. «Av.A nAv.A m

«ow. Ammedo. mwmovo. o~v.~ awe.“ Hun. omeAo. vaovo. mwv.~ va.~ gem. omovoo. vmvovo. mov.H voe.~ N

non. Nomon. moemmo. vov.A mov.A con. mmmAAc. mvmmno. Acv.A Nov.A noA. NoAvoo. mAeovc. man.A mmm.A A

eve. moomao. vwwwvo. oww.~ own. mom. OBNOAO. vnavvo. awn.“ an.~ and. vvnmoo. nmoch. nmm.~ nwm.~ 0A

one. mwewAc. meonvo. omw.A onw.A om». omvao. Nnnmvo. New.A men.A mNA. Acomoc. vnAAvo. vAm.A nAm.A m

oev. nwmwao. wnwmvo. mun.“ own.” wen. mnmvdo. mmomvo. mmo.~ con.H QNA. omomoo. Auwovc. mmv.~ Gav.“ m

o~v. m~nnAo. vamAeo. ANA.A wNu.A own. nNner. mNNNvo. amc.A amo.A AmA. unavoc. avmcvo. mmv.~ vmv.A n

Nae. emHhAo. omoHvo. mn©.~ who.a aNn. wvonAo. Aneuvo. m~o.~ mHo.A cad. mvovoc. vmoovo. one.“ onv.~ 0 wow
woe. vowvo. omwmnc. o~o.A o~o.A mAn.. mnmNAo. onAvc. vem.A enm.A woA. Neneco. oAAceo. wmv.~ mmv.~ m

woe. Anovo. ~mwmmo. cnm.A cnm.A mom. mn-Ao. oooovo. Anm.A Anm.A wcA. mwnvcc. mvvovo. nee.~ Nev.A e

oov. mvmon.. omwmno. aAm.A on.A emu. BNmAAo. «momma. cmv.A ome.A moA. mAAvco. vAmmno. m~v.A va.A n

mam. emomAc. Nunmmo. mmv.A mmv.A emu. NoeAAo. emomno. mev.A mvv.A ooA. mnovoc. «mnovc. vAe.A vAv.A N

man. momon. cmwmno. man.~ mam.A mom. moNNAo. onoovo. eov.A vov.A AoA. nooeoo. nANovc. oov.A mam.~ A

«Ne. AAAon. mAmmvc. mnA.N AAA.N «an. AmonAc. HvNAvo. mnm.~ uno.A NNA. onmoc. mNAAvo. nmm.u mmm.A oA

Ame. msvnuo. mvavo. Amo.n awo.~ mwn. nomnAo. mnmAvo. eAm.A mAm.A mAA. unavoc. mmoAvo. ven.~ nsm.A m

«Ne. omvnAo. nmnAvo. voo.~ eoo.~ eAm. momNAo. nomevo. www.A mmm.A mAA. «havoc. mNcAec. Amm.A Amm.A w

NAv. nvmvo. «moAvo. A~m.A Amm.A eon. owwon. «muons. «an.A eme.A mAA. nvwvoc. osmoeo. onm.A mum.A n

oAq. oAnvo. oooovo. nnw.A omw.A man. cemAAo. nmmmno. Ann.A onn.A nAA. owovoc. owwovo. AAm.A Adm.A 0 *mm
man. mnnon. ovwmmo. nmn.A va.A mmN. AwnAAo. whammo. soo.A noo.A AAA. eomvoo. Nvmovo. ww¢.A mav.~ m

Nan. NAvao. «comma. Ano.A cho.A AmN. onAAo. «enamo. voo.A moo.A moA. mANvoo. oouoeo. moe.A nov.A v

mum. mQNVAO. nvawmo. vmm.~ nam.~ nmw. wNoAHo. evommo. 0mm." omm.A wmo. mmamoo. smoovo. mvv.u ovv.~ m

man. umumao. wawmo. Hom.u oom.~ Na~. auvfimo. wvommo. awe." Nev.A mac. Aawmoo. nmoovo. v~v.~ NNv.~ N

on. Anecdo. «moovo. new.“ new.” mom. HmeHo. womamo. mam.“ mmm.~ woo. mmmmoo. mnnmmo. oov.~ ch.~ A

N: . H500 . COP—A . H.500 . CO‘Q N: . “:00 . COSG . HGOU . G03.“ Np— . 9:09 . =0£m HCOU GOA—m COMHGFQCOU 3 UGO MQW

ooeaaue> use: ooeeAue> cue: ooeeAue> cue:
. . newuuomoum
v. m. A. xuflAwneuwuez

 

.uefiue: o>wueuu0mm< Home: o=Ae> AuwAflaeuAuo: mo omeecu can use .moocafiue> use memo: ownxuocoo ecu owmxuecogm on» we omeego onhuu.~.~ oApeh

62

Estimated Accumulation of Inbreeding_

 

Wright (1931) pointed out that when population size is restricted,
nonrandom mating will occur although it is not intended and this will
cause an accumulation of inbreeding, or average likeness by descent of
the two alleles at a given locus.

In this study the breeding size was restricted to 60 females
and 6 males in each generation. Wright (1931) derived the equation

used to predict inbreeding:

F = F' + Nm * Nf (l + F" - 2 F')

8 Nm Nf

where F is the expected inbreeding when F' was the inbreeding of the
previous generation and F" was the inbreeding two generations previous.

The numbers of breeding males and females are Nm and Nf.

This equation was used to estimate inbreeding each generation.
The results are presented in Table 3. The expected value of inbreeding
was 19% at the end of selection and the average increase each genera-
tion was 2%. The amount of inbreeding was increased, but the response
to selection as the generation progressed showed no inbreeding
depression. This was expected because in this simulation study the
trait was determined only by additive effects of genes. Kempthorne

(1957) has shown that inbreeding depression does not occur without

dominance.

63

Table 3.--Percent Estimated Inbreeding by Generation.

 

 

Generation % F Estimated
Base 0
1 0
2 2
3 5
4 6
5 9
6 11
7 l3
8 15
9 17

10 19

 

64

The Accuracy of the Prediction
Offithe Phenotypic Mean

 

 

The observed and expected phenotypic means are shown in
Table 4 and in Figures 16, 17, and 18. The expected phenotypic means
were calculated by using the formula presented by Falconer (1960b),
R = nihzop, where R is response to selection, n is generation number,

i is selection intensity, h2 is heritability and O is phenotypic

P
standard deviation. The phenotypic mean in generation n was estimated
as Pn and Pn = PC + R, where Po is phenotypic mean in base generation.

The selection intensity was obtained from the table presented
by Nanson (1967) for effective population size less than 50. By
interpolation, the selection intensities were 1.19, .71 and .39 for
25%, 50% and 75% proportion selected for mating, respectively.

The results showed that for the three levels of truncation the
observed and the expected responses agree well throughout the genera-
tions when the heritability value is .1, but when the heritability
values are .3 and .4 the discrepancies between the observed and
expected responses began to occur after the 5th generation of selection
with the expected values greater than the Observed values, except for
the population with 75% proportion selected where the observed and
estimated responses agree well. The larger discrepancies were observed

for the heritability .4 than for heritability .3. This was similar

to results reported by Young (1966).

Estimation of Realized Heritability
Realized heritability was estimated using the method presented

by Hill (1972), where the equation for the estimation was:

65

 

wouovow :ofiuuomonm

HHO.H moo.“ mwN.H mmN.H mvo.N mmm.~ ca
Nwm.H Nwm.~ HvN.H nan.~ Hum.a wa.H m
vem.~ oom.~ wmo.n ebo.H ocm.~ wa.m w
oem.H mmm.H omo.H Nmo.fi wNw.H NNN.H N
Nam.H m~m.~ MHO.H wmm.H umN.H HHN.H o m.
vmv.H mmv.a Hum.a Hem.“ owo.H mmo.H m
one.~ VNQ.H wNm.H fiNm.H VHO.” 0am.~ e
nev.a va.a mmv.~ Nw¢.H Mem.n omm.H m
va.H oNv.~ mvv.H mvv.H Hue.“ wov.~ N
ocv.H oov.~ ocv.H wan.H oo¢.~ cov.~ a
onv.H nov.~ wNm.H o~m.~ vHo.H ooo.~ OH
Noe.H vmv.~ mam.” Nom.H cam.a w~m.~ m
mmv.H ovv.H mmv.a omv.m Nom.H Nmm.H m
Nev.a ove.n mw¢.H va.H mvm.H wNm.H N
mmv.~ mm¢.H HNv.H mov.~ m~m.~ wom.~ o H.
va.H cmv.~ nmv.fi omv.H mav.m wwv.~ m
VNv.H NNv.H NVV.H vwv.H Hue.“ Nov.~ v
o~v.~ v~v.~ wNv.H va.H New.” mvv.~ m
wov.~ wov.~ v~v.~ oHv.H eNv.H va.H N
oov.~ wan.~ oov.H mam.u oov.~ cov.~ H
BHUOQXM Bgvmno VouUOQKm Baumn—O EHUQQKN EEOmn—O
*mn wow me cowumuonoo Suwawpeuwhor

 

.wefluez seemed Home: made: owmxuoeoga wouoomxm use vo>noma nu.v Gamay

66

 

 

wouovom cowpnomonm

Awo.A mmo.A Ham.A mmw.A NmN.N va.N oA
ome.A NNo.A vmw.~ Now.A NoA.N Noo.N m
on.A Aoo.A me.A NON.A coo.N wwa.A w
Nwm.A ohm.A AvN.A MAN.H Ana.A wom.A N
omm.~ evm.~ vwo.A ooo.~ onw.~ Amw.A c e.
mNm.A oAm.A NNc.A AAO.A AmN.A men.~ m
ewe.“ owe.A onm.A vom.A mwo.A moo.A v
Nov.A mmv.A «Am.A cam.A omm.H mum.A n
Ane.A va.~ nmv.A mmv.A mmv.~ omv.A N
oov.A on.A oov.A Noe.A oov.A wan.~ A
noncomxm pO>uomno vouoomxm vo>uomno wouoomxm vo>nomno
Ame *em emm eoAueAeeeo AGAAAeeeAAez

 

.eeeeAeeee--.e eAeee

Means (kg)

2.3

2.2

2.1

2.0 .

1.9

1.8 -

1.7

1.6 .

1.5

1.4

1.3

 

 

67

 

 

A /
)‘I
[P
. ( [I h2=.3
I ’ 0
I
I
/ ’1’
/ ,7 °
’ I
I
I Q / '5
I
I
I
I’ ’
I
. f/ ’
I
[I I, .
I ’1 2
I 7 _ , h =.l
1 I
/ ,l’
- [I
. I, / "
,iy Observed
-_-....._. Expected
l 2 3 4 5 6 7 8 9 10
GENERATION
Fig. 16: Observed and expected phenotypic mean

with random mating and b=25Z

Means (kg)

2.1

2.0

1.9

1.8

1.7

1.6

1.5

1.4

1.3

68

I” ’, 1‘2: 3
fl I 1”
I o /".
i,
I/«.
’ /’.
I
II: It
/ l
I ' / ’
/
/ / ,0 h2=.l
v I, ’ T)

Observed

._ .. ...... Expected

 

12 3 45 678 9 10

GENERATION

Fig. 17: Observed and expected phenotypic mean
with random mating and b=50Z

69

 

 

2.1 v
2.0 II
1.9 II
1.8 ,_
’63
(’5 1.7 I“
2 ,2. h2=.4
3'3 X ..
a/ 3 2
1 6 ,r , h -.3
o "I' I .1 ’ 1
z’ 0 °
” '
I J //
1.5 A
l *1 ,
A.“ ”’..h2=°1
/' - - 3 J
1.4 II 3" 1' Observed
--_._._-_ Expected
1.3 . ,2 . I g. .

 

GENERATION

Fig. 18: Observed and expected phenotypic mean
with random mating and b=751

70

t _ _ t _2
-S) (Xi-XVII (S.-S)

i=0 1

In this equation bc is realized heritability estimate, Si is
the cumulative selection differential to generation 1, Xi is the mean
response at generation i, and S and X are the means of the Si and Xi
over i=0, ...., t generations.

The variance of the estimated value was calculated as:

2(3t+4) 3,2
-2 d
5 s (t+l)(t+2)

t t
where u (be) = h IE}: (xi-i)2 - 66 1:0 (Si-muffl /Et-1) 130 (81-92:,

Gd - G [be (l-bc)/N + bc N]

and t is generation number.

 

v (bc) = u (bc) +

 

 

N=r1+ 1..-1
L4 Nm 4 Nf’J
H-l
M a 7' l + 1
L4 Mm 4 Mf’J

 

 

Mm and Mf is the number of males and females recorded and Nm and Nf is
the number of males and females selected for mating.

The results of the estimated realized heritability by this
formula are presented in Table 5. The estimated realized heritability
values were smaller than the values estimated by variance components
for this simulation in all cases (Table 2.1). The differences can be

explained as in this study the value of the heritabilities were changed

71

Table 5.--The Estimated Value of Realized Heritabilities.

 

 

h2 23% 50% 75%

.1 .08 i .003 .03 i .005 .08 i .009
.3 .24 i .007 .24 i .009 .26 i .020
.4 .33 i .008 .32 i .011 .34 i .019

 

due to the change in the genetic and phenotypic variance. Hill (1972)
discussed the theory of realized heritabilities, but assumed that the
heritability does not change during the experiment. The heritability
did change in this study, and the estimated realized heritability values
differ from the heritability values computed from components. However,
near the beginning of selection, the differences were small and can be

neglected.

‘I‘llll lillll‘l'll‘ill'll.‘

 

 

SUWARY AND CONCLUSIONS

The effect of degree of truncation selection and level of
heritability in simulated populations of poultry subjected to selection
was studied. The trait simulated was body weight at sexual maturity
of native chickens in Indonesia.

The CDC 6500 computer at MSU was used to generate the popu-
lations.

The design of the experiment was 32 factorial. Populations
were simulated for all possible combinations of three levels of
heritability (.l, .3 and .4) and three intensities of selection by
upper truncation of the distribution of phenotypes (25%, 50% and 75%).
The parameters used for mean and standard deviation of body weight were
1.4 kg and .2 kg and the only genetic variance contributed to the
variability was the additive variance.

The size of offspring population in each generation was 600,
300 males and 300 females, with random mortality assumed to occur so
that breeding size after truncation was constant at 6 males and 60
females. Selection was performed for ten generations. For each
parameter combination, 100 replications (populatiOns) were simulated.
Two types of mating systems, random and assortative, were simulated.

The results from random mating and assortative mating showed

the same pattern. The test of difference in slopes of regressions of

72

73

the phenotypic means on generation number for the two types Of mating
was statistically significant in most cases but the difference was
biologically trivial. A small advantage was noted for assortative
mating when selection intensity and heritability were high.

Under the same level of truncation, the response to selection
is proportional to heritability. The best response to selection was
obtained in population with heritability = .4 and the proportion
selected for mating was 25%. The least response to selection was for
population with heritability = .l and proportion selected was 75%.

Discrepancies between Observed and expected response were
noted after the fifth generation of selection when heritability was
.3 or .4, with the expected value greater than the observed value.

No inbreeding depression was observed but the estimated accumu-
lation of inbreeding due to restricted population size averaged 2%
each generation. The heritability values increased somewhat over time
and increased more with assortative mating than with random mating
when 75% or 50% of a population was saved for breeding, but when 25%
was saved increases were similar for both mating systems.

The estimated realized heritability calculated with the formula
presented by Hill (1972) were smaller than the values estimated by
variance components in all cases, because the genetic and phenotypic
variances changed over time in this study.

Assortative mating should not be preferred over random mating,
as the small differences in results are not biologically meaningful.

Also, in Indonesian villages the native chickens are not

raised in confinement and cannot be without major changes in

74

traditional practice. Therefore, only selection with random mating
should be taken under practical consideration.

Although selection of only 25% of the population should pro-
duce greater improvement than selection of 50% or 75% for any level
of heritability, in practice such strong selection should not be
recommended.

There are several reasons, but most importantly it will be
difficult to encourage the village people to save only 25% of the
population.

Disease control is still minimal, natural selection still
plays a significant role in mortality, and the risk of losing an
entire population will be greater than if the population saved were
bigger. Saving 50% to 75% of the population for breeding should be
recommended.

Saving as much as 75% of the population should be preferable
for villages which maintain relatively small populations.

To apply this study in the field a preliminary survey should
be taken to estimate parameters of the Indonesian environment and to
determine other conditions that might alter the results predicted by
this study. Selection works better if the environmental conditions
permit the quantitative trait to be fully expressed. The scarcity of
food, lack of disease control and lack of cooperation of rural peOple
are a few among many factors that may hinder potential improvement.

It is important to study the characteristics of the native
chickens as they have adapted to village conditions and developed

great resistance to disease. The idea of introducing a "new" breed

75

(exotic breed) to the villages is still questionable, primarily

because of the questionable adaptability to the Indonesian environment.
This study has shown that without introducing any blood from

other breeds, selection along with random mating within the native

chicken population can bring a significant improvement.

III III I! IIII .111. I‘ll I. III I II III III III. 1‘ 9' illIIIIIll.llIIIl-Ill lil‘l IIII I!

 

LITERATURE CITED

LITERATURE CITED

Bereskin, B., C. E. Shelby, K. E. Rowe, W. E. Urban, Jr., C. T. Blunn,
A. B. Chapman, V. A. Garwood, L. N. Hazel, J. F. Lasley, W. T.
Magee, J. W. McCarty, and J. A. Whatley, Jr. 1968. Inbreed-
ing and swine productivity traits. J. Animal Sci. 27:339.

Bereskin, B., C. E. Shelby, and L. N. Hazel. 1969. Monte Carlo
studies of selection and inbreeding in swine. I. Genetic and
phenotypic trends. J. Animal Sci. 29:678.

Bereskin, B. 1972. Monte Carlo studies of selection and inbreeding
in swine. IV. Selection response. J. Animal Sci. 34:726.

Berruecos, J. M., E. U. Dill, and O. W. Robison. 1970. Selection
for low backfat thickness in swine. J. Animal Sci. 30:844.

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.

Chai, C. K. 1956. Analysis of quantitative inheritance of body size
in mice. 11. Gene action and segregation. Genetics 41:165.

Chapman, A. B. 1951. Effectiveness of selection in laboratory animals.
J. Animal Sci. 10:3.

Dalton, D. C. and T. L. Bywater. 1963. The effect of selection for
litter size and litter weight at weaning in mice maintained
on two diets. Animal Production 5:317.

Dalton, D. C. 1967. Selection for growth in mice on two diets.
Animal Production 9:425.

Dinkel, C. A., D. A. Busch, J. A. Minyard, and W. R. Trevillyan. 1968.
Effects of inbreeding on growth and conformation of beef
cattle. J. Animal Sci. 27:313.

Eisen, E. J. 1975. Population size and selection intensity effects
on long term selection response in mice. Genetics 79:305.

Enfield, F. D., R. E. Comstock, and O. Braskerud. 1966. Selection
for pupa weight in Tribolium castaneum. 1. Parameters in
base populations. Genetics 54:523.

76

77

Falconer, D. S. 1960. Introduction to quantitative genetics. The
Ronald Press Company, New York.

Frankham, R., L. P. Jones, and J. S. F. Barker. 1968. The effect of
population size and selection intensity in selection for a
quantitative character in Drosophila. 1. Short term response
to selection. Genetical Research 12:237.

Fraser, A. S. 1957a. Simulation of genetic system by automatic
digital computers. I. Introduction. Aust. J. Biol. Sci.
10:484.

Fraser, A. S. 1957b. Simulation of genetic systems by automatic
digital computers. II. Effects of linkage on rates of advance
under selection. Aust. J. Biol. Sci. 10:492.

Gill, J. L. 1965a. Effects of finite size on selection advance in
simulated genetic populations. Aust. J. Biol. Sci. 18:599.

Gill, J. L. 1965b. A Monte Carlo evaluation of predicted selection
response. Aust. J. Biol. Sci. 18:999.

Gill, J. L. 1965c. Selection and linkage in simulated genetic popu-
lations. Aust. J. Biol. Sci. 18:1171.

Gill, J. L. and B. A. Clemmer. 1966. Effects of selection on degree
of inbreeding. Aust. J. Biol. Sci. 19:307.

Godfrey, G. F. and B. L. Goodman. 1955. Selection for small and
large body size in broiler chickens. Poultry Sci. 34:1196.
(Abstr.)

Hanrahan, J. P., E. J. Eisen, and J. E. Legates. 1973. Effects of
population size and selection intensity on short term response
to selection for post-weaning gain in mice. Genetics 73:513.

Hill, W. G. 1972. Estimation of realized heritabilities from
selection experiments. II. Selection in one direction.
Biometrics 28:767.

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.

Kemp, K. E. and W. T. Magee. 1966. Genetic changes in a simulated
crossbred swine herd. J. Anim. Sci. 25:1271. (Abstr.)

Kempthorne, O. 1957. An introduction to genetic statistics. John
Wiley and Sons, Inc., New York.

Kyle, W. H. and A. B. Chapman. 1953. Experiment check of the effective-
ness of selection for a quantitative character. Genetics 38:421.

III ‘II I .II..\.'IIl

78

Li, C. C. 1955. Population genetics. The University of Chicago
Press, Chicago, Ill.

Lush, J. L. 1948. The genetics of population (Mimeo). Dept. of
Animal Science, Iowa State University, Ames, Iowa.

Lush, J. L. 1956. Animal breeding plans. Iowa State College Press,
Ames, Iowa.

Martin, G. A., E. W. Glazener, and W. L. Blow. 1953. Efficiency of
selection for broiler growth at various ages. Poultry Sci.
32:716.

MacArthur. 1949. Selection for small and large body size in the
house mouse. Genetics 34:194.

MacLaury, D. W. and A. W. Nordskog. 1956. Effects of inbreeding on
mortality in the domestic fowl. Poultry Sci. 35:582.

McBride, G. and A. Robertson. 1963. Selection using assortative
mating in Drosophila melanogaster. Genet. Res. 4:356.

Miller, R. H., J. E. Legates, and C. C. Cockerham. 1963. Estimation
of non-additive hereditary variance in traits of mice.
Genetics 48:177.

Nanson, A. 1967. Table de la differentielle de Sélection dans la
distribution normale (0, l). Biometric Praximetrie 8:40.

Nelms, G. E. and P. O. Stratton. 1967. Selection practiced and
phenotypic change in a closed line of beef cattle. J. Animal
Sci. 26:274.

Nishida, A. and T. Abe. 1974. The distribution Of genetic and
environmental effects and the linearity of heritability.
Canadian J. Genetics and Cytology 16:3.

Parker, R. J. 1966. Genetic correlation and response to selection in
simulated populations. Ph.D. Thesis, Michigan State Univer-
sity, East Lansing, Michigan.

Rahnefeld, G. R., w. J. Boylan, R. E. Comstock, and M. Singh. 1963.
Mass selection for postweaning growth in mice. Genetics
48:1567.

Rollins, W. C., S. W. Mead, W. M. Regan, and P. W. Gregory. 1949.
The effect of inbreeding upon growth in Jersey cattle.
Genetics 34:724.

Scheinberg, E. 1968. Methodology of computer genetics research.
Canadian J. Genetics and Cytology 10:754.

79

Schoffner, R. N. 1948. The reaction of the fowl to inbreeding.
Poultry Sci. 27:448.

Siegel, P. B. 1962. A double selection experiment for body weight
and breast angle at eight weeks of age in chickens. Genetics
47:1313.

Sutherland, T. M., P. E. Biondini, and L. H. Haverland. 1968.
Selection under assortative mating in mice. Genet. Res. 11:
171.

Thomas, C. H., W. L. Blow, C. 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.

Von Krosigk, C. M. and J. L. Lush. 1958. Effect of inbreeding on
production in Holsteins. Journal Dairy Sci. 41:105.

Wilson, 8. P., W. H. Kyle, and A. E. Bell. 1965. The effects of
mating systems and selection on pupa weight in Tribolium.
Genet. Res. 6:341.

Wright, S. 1931. Evolution in Mendelian populations. Genetics 16:97.

Yao, T. S. 1959. Additive and dominance effects of genes in egg
production 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.

Young, S. S. Y. 1966. Computer simulation of directional selection
in large populations. 1. The programme, the additive and the
dominance models. Genetics 53:189.

MICHIGAN STATE UNIVERSITY LI

03

3 1293

BRARms
4

082 470