Abstract
Human genetic variation is the incarnation of diverse evolutionary history, which reflects both selectively advantageous and selectively neutral change. In this study, we catalogue structural and functional features of proteins that restrain genetic variation leading to single amino acid substitutions. Our variation dataset is divided into three categories: i) Mendelian disease-related variants, ii) neutral polymorphisms and iii) cancer somatic mutations. We characterize structural environments of the amino acid variants by the following properties: i) side-chain solvent accessibility, ii) main-chain secondary structure, and iii) hydrogen bonds from a side chain to a main chain or other side chains. To address functional restraints, amino acid substitutions in proteins are examined to see whether they are located at functionally important sites involved in protein-protein interactions, protein-ligand interactions or catalytic activity of enzymes. We also measure the likelihood of amino acid substitutions and the degree of residue conservation where variants occur. We show that various types of variants are under different degrees of structural and functional restraints, which affect their occurrence in human proteome.
Introduction
Every human genome contains thousands of single amino acid variations (SAVs). While many are "neutral" (having no noticeable effect), others cause devastating Mendelian diseases or drive cancer. The challenge lies in prediction: how do we distinguish a harmless "passenger" mutation from a "driver" of disease? This study addresses this by investigating the structural and functional "restraints" that evolution places on proteins. The location of a mutation—whether it's on the protein's surface, buried in its core, or part of a catalytic site—determines its likelihood of being pathogenic. This paper performs a large-scale census to quantify these environmental preferences across the human proteome.
Methods and Datasets
The researchers employed a multi-step computational pipeline to map sequence-level variations onto known 3D structures in the Protein Data Bank (PDB). The study utilized four primary datasets:
- SVD (Swiss-Prot Variant Disease): 16,776 Mendelian disease-related variants.
- SVP (Swiss-Prot Variant Polymorphism): 32,748 neutral polymorphisms.
- SAP (Ensembl Single Amino acid Polymorphism): 29,541 verified neutral SNPs.
- CSM (COSMIC): 5,260 somatic mutations found in cancer.
The local environment of each variant was characterized using JOY to analyze solvent accessibility, secondary structure, and hydrogen-bonding patterns. Substitution scores were calculated using BLOSUM62 and ESST (Environment Specific Substitution Tables), while evolutionary conservation was measured via Shannon entropy. Finally, variants were checked against known functional sites, such as metal-binding regions and catalytic active sites.
Results: Structural and Functional Restraints
The analysis revealed that disease-causing changes are strictly restrained by protein architecture, preferring buried, highly conserved, and functionally critical sites that neutral variations typically avoid.
Structural Localization and the "Buried" Penalty
Mendelian disease variants (SVD) are twice as likely to occur in the solvent-inaccessible core of a protein compared to neutral variants. Specifically, 42.25% of SVDs occur in buried regions, compared to only ~19% for neutral variants (SVP/SAP). Mutations in the core often destabilize the entire protein fold, whereas surface mutations are more easily tolerated. Somatic cancer mutations (CSM) occupy a middle ground, being more likely to be buried than neutral SNPs but less likely than Mendelian disease variants.
Torsion and Hydrogen Bonding
Specific structural features act as strong restraints. Residues with a "positive $\phi$ angle" are highly sensitive; mutations here are almost always deleterious because only specific amino acids (like Glycine) can comfortably adopt this geometry. Additionally, disease-related variants occur significantly more often at residues that form side-chain hydrogen bonds. Breaking these bonds often compromises structural integrity or functional tuning.
Functional Site Enrichment
SVDs are highly enriched at functional sites. For example, ~10.4% of SVDs occur at metal-binding or general binding sites, while neutral variants occur at these sites less than 1% of the time. This confirms that mutations disrupting direct biochemical activity are far more likely to result in clinical disease.
Substitution Tolerance and Conservation
SVD and CSM variants have significantly lower (more negative) ESST scores than neutral variants, particularly in buried regions. This indicates that disease mutations frequently involve "radical" changes (e.g., swapping a small hydrophobic residue for a large charged one) that the local environment cannot accommodate. Furthermore, SVDs occur at sites with significantly higher conservation (lower Shannon entropy) than neutral variants, confirming that disease hits the "parts that haven't changed in millions of years."
Conclusion
By mapping thousands of human genetic variations onto 3D protein structures, this study establishes quantitative rules for variant pathogenicity. The clear distinction in structural and functional environments between disease-related and neutral variations provides a powerful framework for future predictive algorithms. Integrating these structural restraints into personalized medicine could significantly improve our ability to identify high-risk mutations in clinical genomics.
Summary updated