Proteome-wide model for human disease genetics.

Area of Science:

• Genomics
• Computational Biology
• Human Genetics

Background:

• Missense variants are challenging to interpret for disease causation, hindering genetic diagnosis and therapeutic development.
• Existing computational prediction methods lack proteome-wide calibration, limiting their clinical utility for variants in novel genes.
• Accurate prediction of missense variant effects is crucial for understanding genetic disorders.

Purpose of the Study:

• To develop a deep generative model, popEVE, for accurate, proteome-wide prediction of missense variant pathogenicity.
• To apply popEVE to identify novel candidate genes associated with developmental disorders.
• To demonstrate the utility of popEVE in genetic analysis, particularly for rare diseases.

Main Methods:

• Developed popEVE, a deep generative model integrating evolutionary and population sequence data.
• Evaluated popEVE's performance on proteome-wide prediction tasks.
• Applied popEVE to a developmental disorder cohort to identify candidate disease genes.

Main Results:

• popEVE achieves state-of-the-art performance in predicting variant effects without overestimating deleterious variants.
• Identified 442 candidate genes in a developmental disorder cohort, including 123 novel candidates.
• Candidate genes showed functional similarity to known disease genes, and variants were located in critical regions.

Conclusions:

• popEVE offers a robust solution for interpreting missense variants across the proteome.
• The model successfully identifies novel candidate genes for developmental disorders, even without cohort-wide enrichment.
• popEVE provides a powerful new approach for genetic analysis, especially for rare diseases, using patient exomes alone.