AlphaMissense for Identifying Pathogenic Missense Mutations in DNA Damage Repair Genes in Cancer

Area of Science:

• Genomics and Bioinformatics
• Cancer Genetics
• Artificial Intelligence in Medicine

Background:

• Missense variants in DNA damage repair (DDR) genes are crucial in cancer development.
• Accurate pathogenicity prediction of these variants is essential for clinical decision-making.
• AlphaMissense is an AI tool for predicting missense variant pathogenicity.

Purpose of the Study:

• To evaluate the accuracy of AlphaMissense predictions for DDR gene variants.
• To assess the utility of AlphaMissense in identifying novel pathogenic mutations.
• To correlate AlphaMissense predictions with genomic and clinical characteristics in cancer patients.

Main Methods:

• Analysis of sequencing data from 56,965 cancer patients (MSK-IMPACT) and TCGA/PCAWG cohorts.
• Evaluation of AlphaMissense pathogenicity predictions in six common DDR genes.
• Classification of missense mutations as known pathogenic, newly identified pathogenic by AlphaMissense, or benign.

Main Results:

• AlphaMissense identified 17.5% of unique DDR gene missense mutations as newly pathogenic in the MSK-IMPACT cohort.
• Tumors with newly identified pathogenic BRCA1/2, PALB2, and RAD51C mutations showed increased homologous recombination deficiency signatures.
• New pathogenic ATM mutations correlated with fewer TP53 mutations and improved tumor control post-irradiation.

Conclusions:

• AlphaMissense shows promise in identifying previously unrecognized pathogenic DDR gene mutations.
• The accuracy of AlphaMissense predictions is gene-dependent.
• Clinical and functional validation are necessary to confirm AlphaMissense predictions for broader application.