Off-campus UMass Amherst users: To download campus access dissertations, please use the following link to log into our proxy server with your UMass Amherst user name and password.
Non-UMass Amherst users: Please talk to your librarian about requesting this dissertation through interlibrary loan.
Dissertations that have an embargo placed on them will not be available to anyone until the embargo expires.
Author ORCID Identifier
https://orcid.org/0000-0001-9139-0152
AccessType
Campus-Only Access for One (1) Year
Document Type
dissertation
Degree Name
Doctor of Philosophy (PhD)
Degree Program
Chemistry
Year Degree Awarded
2023
Month Degree Awarded
September
First Advisor
Jianhan Chen
Subject Categories
Biophysics
Abstract
Quantitative and predictive models of protein structure-function relationships is crucial for diagnosing the effect of perturbations like mutations on human disease. While there is often abundant functional data describing the impact of mutations at the cellular level, and high-resolution structural data the protein’s wild-type lowest-energy fold, the atomic-level consequences on the structure and dynamics of a protein mutation are challenging to obtain. Physics-based simulations can bridge this gap by describing the native conformational ensemble as well as the effects of functional perturbations with atomistic resolution. Where simulations are infeasible, mixed statistical/empirical potentials can quantify the effects of mutations without the need for expensive sampling. In the first three chapters we used free energy calculations to quantify the free energy of hydrophobic dewetting of protein channels and model nanopores. We first developed a protocol to quantify the effects on hydration free energy of pore geometries and cosolvents in protein-like nanopores. Subsequently, we quantified the hydration free energy in the BK channel and demonstrated that the hydration of the closed pore is a key thermodynamic step in the gating transition for the channel and its mutants. Finally, we demonstrated that the molecular mechanism of the BK channel activating small molecule arises not by specific binding, but by directly modulating the hydration of the pore. In the final two chapters, we built predictive sequence-to-function models of two proteins, the BK channel and Hsp70 chaperone, by leveraging machine learning to train physics-based descriptors to reproduce experiments. In chapter five we used MD simulations to describe the interactions of all amino acids with Hsp70, determined an optimal reweighting these interaction terms to reproduce binding affinity data, and found that the model predicts the proper binding register and orientation crystallographic structures. In chapter six we describe the construction of a model of BK channel voltage gating by training physics-based description of the impacts of mutation to reproduce over 450 experimentally-determined point-mutations. Importantly, the model identified four novel mutations key to voltage gating. All together this work demonstrates the importance of physics-based modeling and simulation in the general understanding and predictive modeling of protein function.
DOI
https://doi.org/10.7275/35733995
Recommended Citation
Nordquist, Erik B., "Understanding Protein Structure-Function Relationships Using Simulations and Machine Learning" (2023). Doctoral Dissertations. 2934.
https://doi.org/10.7275/35733995
https://scholarworks.umass.edu/dissertations_2/2934
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.