Development of molecular dynamics force field of YOPRO-1 and deep learning models for protein classification
Cyanine dyes, such as Oxazole yellow (YOPRO), are almost non-fluorescent in water but their fluorescence is greatly enhanced after intercalation in double-stranded DNA, providing the basis of DNA concentration assays. The rationale for this property is the flexibility difference of the conformations of the molecule in different environments, mainly attributed to the linker dihedral rotations. We compared two methods for deriving the specific dihedral force field on the linker of YOPRO, namely by modifying the AMBER generated force field (GAFF) and by using the IPolQ fitting protocol. There are two dihedral angles and the IPolQ method showed that their potential surfaces are coupled. Thus, going beyond the GAFF approach, coupled dihedral surfaces were obtained for the ground S0 and first excited S1 electronic states. Molecular Dynamics (MD) simulations of YOPRO were carried out in water and intercalations using these force field models. The MD simulations started at the minima of the S0 state vertically excited to the S1 state. The contrast between YOPRO conformational relaxation on the S1 surface in water and when intercalated provided the non-radiative relaxation pathways relevant to fluorescence decay and explain the differences in quantum yield. For the second topic, we investigated a number of deep machine learning (ML) models for protein family classification. We used one dimensional sequence and three dimensional secondary structural information of proteins as the input for training the neural network models. The results show that deeper convolutional networks of the Long Short Term Memory (LSTM) variety significantly enhanced the prediction accuracies compared to less sophisticated models. The addition of the secondary structural information greatly increases the testing accuracies with the training data size remaining the same. Proteins belonging to different families can be successfully distinguished using these methods.
Read
- In Collections
-
Electronic Theses & Dissertations
- Copyright Status
- In Copyright
- Material Type
-
Theses
- Authors
-
Jin, Chi
- Thesis Advisors
-
Hunt, Katharine C.
Cukier, Robert I.
- Committee Members
-
Hegg, Eric L.
Swain, Greg M.
Hong, Heedeok
- Date Published
-
2019
- Program of Study
-
Chemistry - Doctor of Philosophy
- Degree Level
-
Doctoral
- Language
-
English
- Pages
- xi, 145 pages
- ISBN
-
9781392227510
1392227518
- Permalink
- https://doi.org/doi:10.25335/gfhq-y832