Computational Investigations into Binding Dynamics of Tau Protein Antibodies: Using Machine Learning and Biophysical Models to Build a Better Reality
Date of Completion
Eric May; Yongku Cho; Adam Zweifach
University Scholar Major
Structural Biology and Biophysics
Biophysics | Other Biochemistry, Biophysics, and Structural Biology | Structural Biology
Misregulation of post-translational modifications of microtubule-associated protein tau is implicated in several neurodegenerative diseases including Alzheimer’s disease. Hyperphosphorylation of tau promotes aggregation of tau monomers into filaments which are common in tau-associated pathologies. Therefore, tau is a promising target for therapeutics and diagnostics. Recently, high-affinity, high-specificity single-chain variable fragment (scFv) antibodies against pThr-231 tau were generated and the most promising variant (scFv 3.24) displayed 20-fold increased binding affinity to pThr-231 tau compared to the wild-type. The scFv 3.24 variant contained five point mutations, and intriguingly none were in the tau binding site. The increased affinity was hypothesized to occur due to allosteric communication between the framework region and binding site. Multi-microsecond all-atom molecular dynamics simulations were conducted for four systems – the wild-type antibody and the mutant, with and without tau. Correlation of All Rotameric and Dynamical States (CARDS) software was used to quantify allostery in terms of mutual information (MI), or the dependence between two variables. The mutant exhibited much higher total MI than the wild-type as well as MI relative to target sites of interest, including the four residues that bind directly to the phosphate group on tau. DiffNets, a supervised autoencoder with a classification task, was used to distinguish the relevant motions separating wildtype from 3.24 mutant. Results showed long-range expansion within the mutant stemming from mutation Ile61. Recent work has been aimed towards quantifying optimal collective variables for discriminating wild-type from mutant ensembles for use in free energy calculations.
Lee, Katherine, "Computational Investigations into Binding Dynamics of Tau Protein Antibodies: Using Machine Learning and Biophysical Models to Build a Better Reality" (2022). University Scholar Projects. 82.