Diverse ancestral representation improves genetic intolerance metrics.

Area of Science:

• Genomics
• Human Genetics
• Bioinformatics

Background:

• Genomic databases are crucial for identifying disease-associated genomic regions.
• Current databases lack sufficient ancestral diversity, limiting the performance of intolerance metrics.
• Key metrics include Residual Variance Intolerance Score (RVIS), Missense Tolerance Ratio (MTR), and Loss-of-Function Observed/Expected ratio (LOF O/E).

Purpose of the Study:

• To evaluate the impact of ancestral diversity on the performance of human genome intolerance metrics.
• To compare the effectiveness of scores trained on diverse ancestries versus those trained on larger, less diverse datasets.
• To assess the saturation point of current intolerance scores based on European ancestry.

Main Methods:

• Analysis of whole-exome sequencing data from UK Biobank (460,551 participants) and gnomAD (125,748 participants) across multiple ancestries.
• Testing RVIS, MTR, and LOF O/E metrics.
• Comparing performance of scores trained on different ancestral groups, particularly African, Admixed American, and European ancestries.

Main Results:

• Increased ancestral representation significantly improves the performance of intolerance metrics compared to sample size alone.
• Scores trained on African and Admixed American data show superior resolution in identifying disease risk genes, including those for neurodevelopmental disorders.
• Missense Tolerance Ratio trained on a smaller multi-ancestry dataset outperformed a much larger European-only dataset, indicating European-based scores may be reaching saturation.

Conclusions:

• Enhanced population representation in genomic resources is essential for improving disease gene discovery and precision medicine.
• Ancestry-specific intolerance scores derived from diverse populations offer greater predictive power.
• Publicly available scores are provided to facilitate research and drug discovery.