Evaluating variant pathogenicity prediction tools to establish African inclusive guidelines for germline genetic...

Area of Science:

• Genomics
• Bioinformatics
• Population Genetics

Background:

• Genetic germline testing is limited for African patients due to a lack of ancestrally relevant genomic data.
• European-biased variant databases and prediction guidelines hinder accurate genetic analysis in diverse populations.
• The performance of variant pathogenicity prediction tools (VPPTs) in African populations remains largely unassessed.

Purpose of the Study:

• To evaluate the performance of 54 VPPTs using genetic data from Southern African and European men with prostate cancer.
• To identify VPPTs that perform optimally across diverse ancestries, particularly for African populations.
• To screen millions of variants of unknown significance for potential functional and oncogenic impact.

Main Methods:

• Assessed 54 VPPTs on 145,291 known pathogenic or benign variants from Southern African and European men.
• Analyzed VPPT performance metrics including sensitivity, specificity, and false positive/negative rates.
• Screened 5.3 million variants of unknown significance using prioritized VPPTs.

Main Results:

• VPPT sensitivity was lower for African data (0.66) compared to European data (0.71).
• Several VPPTs (MetaSVM, CADD, Eigen-raw, BayesDel-noAF, phyloP100way-vertebrate, MVP) performed well across ancestries.
• Specific tools showed ancestry-biased performance (e.g., MutationTaster for African, REVEL for European), but selected workflows narrowed the prediction gap for African data.

Conclusions:

• VPPT sensitivity generally favors European genetic data, highlighting an existing bias.
• The study provides crucial guidelines for selecting VPPTs to enhance rare pathogenic variant prediction in African disease studies.
• These findings aim to improve genetic testing accuracy and outcomes for African patients by addressing data limitations.