Dissertation
Novel computational biophysics methods for unraveling the impact of missense variants on hearing loss
University of Iowa
Doctor of Philosophy (PhD), University of Iowa
Spring 2026
DOI: 10.25820/etd.008474
Abstract
Advancements in experimental protein structure determination have massively expanded the Protein Data Bank (PDB) filling in significant knowledge gaps regarding protein structure and function. However, the PDB is not complete. Experimental structures are often too low resolution for accurate proton placement, protein structures are too large to solve experimentally, and the resources required to experimentally solve all possible structures are not available. Focusing on proteins associated with hearing loss, the most common sensory deficit, only 40% have experimental structural coverage. This gap becomes even larger when including genetic variants or protein-protein interactions. Protein structure and function are tightly coupled. Therefore, it is necessary to fully resolve the impact of acid-base chemistry and genetic variants on protein structure to better understand the resulting function.
Protein acid-base chemistry is especially important for catalytic sites in protein-protein interactions and ligand binding sites. To accurately place protons in protein structures, we can use computational methods to consider the effects of pH on titratable residues (ASP, GLU, HIS, LYS) in protein structure. We introduce a novel method for fast prediction of protein acid-base chemistry using statistical mechanics and side-chain optimization with a polarizable force field and implicit solvent. Our method predicts pKa, a common metric for measuring acid-base chemistry predictors, to a root-mean-squared-error (RMSE) of 0.68. This RMSE is on par with the available deep-learning methods. However, our method is also able to capture the effects of binding partners on proton placement that are left out of the deep-learning methods. With this method, we can apply pH effects to protein structures and protein-protein complexes to better study their structure and function.
We characterize the biophysical effects of genetic variants on protein structures associated with hearing loss. The Deafness Variation Database (DVD) is a public resource of deafness variants, containing over 380,000 missense variants across 224 genes, with 303,577 classified as a variant of uncertain significance (VUS). To address the challenge of interpreting each deafness associated VUS, we evaluate a family of probabilistic frameworks to quantify the strength of computational evidence based on ACMG/AMP recommendations. Our protein folding-informed Bayesian model results in over 28,000 VUSs reaching very strong evidence of pathogenicity with a false positive rate of only 0.14%. From these VUSs, we identify twelve probands where the patient’s genetic diagnosis is upgraded to likely pathogenic/pathogenic. We highlight two variants that cause clear structural disruption, demonstrating the impact of biophysical characterization on variant evaluation.
We combined deep learning protein structure prediction with thermodynamic data to determine disease variants that disrupt protein-protein interactions (PPIs). We collected 295 possible PPIs for DVD genes and modeled them using AlphaFold3. We then calculated binding free energy differences for the 5,999 variants in the interacting regions of confident PPIs. In combination with genetic predictors CADD and REVEL, we compiled a list of ~500 variants that are likely disrupting PPIs. We selected three variants seen in probands to highlight as binding disrupters. Understanding the mechanism of disease is a step towards developing patient-specific diagnosis and treatment methods. While these methods focus on hearing loss, they are not disease specific and can be applied in other contexts.
Details
- Title: Subtitle
- Novel computational biophysics methods for unraveling the impact of missense variants on hearing loss
- Creators
- Rose Arena Gogal
- Contributors
- Michael J. Schnieders (Advisor)Terry A. Braun (Committee Member)Richard J.H. Smith (Committee Member)Claudio Margulis (Committee Member)
- Resource Type
- Dissertation
- Degree Awarded
- Doctor of Philosophy (PhD), University of Iowa
- Degree in
- Biomedical Engineering
- Date degree season
- Spring 2026
- DOI
- 10.25820/etd.008474
- Publisher
- University of Iowa
- Number of pages
- xvi, 161 pages
- Copyright
- Copyright 2026 Rose Gogal
- Language
- English
- Date submitted
- 04/10/2026
- Description illustrations
- Illustrations, graphs, charts, tables
- Description bibliographic
- Includes bibliographical references (pages 129-144).
- Public Abstract (ETD)
Protein structure and function are closely tied to human health. Investigating all proteins and the naturally occurring changes to their native sequence with wet-lab experiments is costly and difficult. In this work, we develop computational algorithms to study acid-base chemistry in protein structure, and the effect of a single residue change in protein sequence on the protein structure and its ability to interact with other proteins.
The acid-base chemistry of proteins is fundamental to their function. Current computational methods to study acid-base chemistry require significant resources or lack the ability to include the protein’s environment. Here, we develop a method that reaches similar accuracy to current methods when comparing to experimentally measured values while reducing computational cost and including protein environment.
We investigate the impacts of single amino acid changes on both protein structure and its interactions specifically in the context of hearing loss. Hearing loss is the most common impairment of the main senses affecting nearly 5% of the world’s population. About 60% of cases diagnosed in infants are attributed to inheritance, but it is difficult to pinpoint the cause due to the vast genetic landscape. We integrate biophysical data that quantifies the effect of a change to protein sequence on both protein folding and protein-protein binding in combination with genetic data to select the changes most likely to cause disease. The results are pipelines used to evaluate the inherited causes of hearing loss. These pipelines are not hearing loss specific and can be applied to other disease contexts.
- Academic Unit
- Roy J. Carver Department of Biomedical Engineering
- Record Identifier
- 9985176872802771
Metrics
1 File views/ downloads
1 Record Views