Annotation of Human Exome Gene Variants with Consensus Pathogenicity

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Accurate annotation of exome variants with phenotypic effects is crucial for understanding genetic diseases.
• Limited empirical data hinders the functional characterization of many genetic variants.

Purpose of the Study:

• To develop a novel predictive annotation method for exome variants lacking empirical data.
• To improve the prioritization of human genetic variants for clinical and biological applications.

Main Methods:

• A stacked ensemble of supervised machine learning models, including distributed random forest and gradient boosting machines, was employed.
• Models were trained and validated using ClinVar database annotations and applied to 84 million non-synonymous single nucleotide variants (SNVs).
• A consensus model integrated 39 functional mutation impacts, cross-species conservation, and a novel gene indispensability score.

Main Results:

• The gene indispensability score significantly enhanced prediction accuracy by considering variant pathogenicity across essential and tolerant genes.
• The consensus model achieved high consistency with input scores while minimizing false predictions.
• Input scores were ranked by predictive ability, aiding interpretation.

Conclusions:

• The developed predictive annotation method offers a robust approach for variant effect prediction in the absence of empirical evidence.
• This tool can aid in prioritizing human exome variants, facilitating clinical diagnostics and biological research.