Back to search

University of Edinburgh

Improving the prediction and prioritization of pathogenic and phenotypically important human variants

Abstract

dc:description.abstract

The interpretation of genetic variation within the human proteome remains a critical challenge in genomic medicine, particularly when considering the diverse biophysical and structural contexts of proteins. This thesis addresses the limitations of current variant interpretation frameworks by systematically investigating the role of intrinsic structural disorder and regional protein features in modulating the pathogenicity of two major variant types: missense substitutions and in-frame insertions and deletions (in-frame indels). In Chapter 1, I provide a comprehensive review of protein structure, disorder, and the mutational mechanisms that give rise to Missense variants and in-frame indels, laying the foundation for a region-aware view of variant tolerance. Chapter 2 presents a large-scale benchmarking of variant effect predictors (VEPs), revealing that their performance varies substantially across proteins with differing structural and functional properties. This work highlights that regions enriched in intrinsic disorder can inflate predictive accuracy due to the overrepresentation of benign variants, raising concerns about the reliability of global performance metrics such as AUROC. Building on these insights, Chapter 3 focuses on intrinsically disordered protein regions (IDRs) and their unique challenges in variant interpretation. Despite their structural flexibility, IDRs are not uniformly tolerant to variation. This chapter uncovers a notable depletion of pathogenic variants in disordered regions, particularly in autosomal recessive genes, yet also identifies specific mechanisms, such as gain-of-function and transcriptional disruption, by which disease-associated variants act within IDRs. Furthermore, I demonstrate that conventional VEPs underperform in these regions, and show that region-specific prediction thresholds can improve diagnostic sensitivity. In Chapter 4, I turn attention to in-frame indels, an understudied class of variants with profound structural implications. Through an integrative analysis combining ClinVar and gnomAD data with structural, biophysical, and disorder annotations, I find that pathogenic in-frame indels are significantly enriched in disordered regions and are generally longer than their benign counterparts. I also demonstrate that in-frame deletions are more frequent and shorter than in-frame insertions, whereas in-frame insertions are more frequently buffered by local flexibility and intramolecular bonds. Finally, I introduce the Indels Intolerance Score (IIS), an empirical observed/expected metric capturing regional intolerance to in-frame indels and validate its ability to distinguish pathogenic from benign in-frame indels. To enable proteome-wide application, I train a machine learning model on 65 structural, evolutionary, and ensemble features, achieving high predictive accuracy (R² = 0.873). This model allows for intolerance estimation even in poorly characterized proteins. Altogether, this thesis advances our understanding of how local protein context, particularly intrinsic disorder, and structural constraints, modulates variant impact. It provides a robust analytical framework for improving variant interpretation across diverse structural landscapes, with practical implications for clinical genomics, molecular diagnostics, and protein engineering and evolution.

Author and committee

dc:creator, dc:contributor.*
Authors dc:creator
  • Fawzy, Mohamed
  • Canela Xandri, Oriol
Advisor dc:contributor.advisor
  • Marsh, Joseph

Subjects

dc:subject × 8

Rights

Language dc:language.iso
en

Identifiers

dc:identifier.*

Chain of custody

source
Harvested from
University of Edinburgh
Base URL
era.ed.ac.uk/server/oai/request
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
citation

Fawzy, Mohamed; Canela Xandri, Oriol. Improving the prediction and prioritization of pathogenic and phenotypically important human variants. 2026. https://era.ed.ac.uk/handle/1842/44813