Localized structural frustration for evaluating the impact of sequence variants

The rapidly declining costs of sequencing human genomes and exomes are providing deeper insights into genomic variation than previously possible. Growing sequence datasets are uncovering large numbers of rare single-nucleotide variants (SNVs) in coding regions, many of which may even be unique to single individuals. The rarity of such variants makes it difficult to use conventional variant-phenotype associations as a means of predicting their potential impacts. As such, protein structures may help to provide the needed means for inferring otherwise difficult-to-discern rare SNV-phenotype associations. Previous efforts have sought to quantify the effects of SNVs on structures by evaluating their impacts on global stability. However, local perturbations can severely impact functionality (such as catalysis, allosteric regulation, interactions and specificity) without strongly disrupting global stability. Here, we describe a workflow in which localized frustration (which quantifies unfavorable residue-residue interactions) is employed as a metric to investigate such effects. We apply frustration to study the impacts of a large number of SNVs available throughout a number of next-generation sequencing datasets. Most of our observations are intuitively consistent: we observe that disease-associated SNVs have a strong proclivity to induce strong changes in localized frustration, and rare variants tend to disrupt local interactions to a larger extent than do common variants. Furthermore, we observe that somatic SNVs associated with oncogenes induce stronger perturbations at the surface, whereas those associated with tumor suppressor genes (TSGs) induce stronger perturbations in the interior. These findings are consistent with the notion that gain-of-function (for oncogenes) and loss-of-function events (for TSGs) may act through changes in regulatory interactions and basic functionality, respectively