Machine learning-based penetrance of genetic variants.

Area of Science:

• Genomics
• Computational Biology
• Precision Medicine

Background:

• Accurate estimation of variant penetrance is essential for precision medicine and genetic risk assessment.
• Traditional methods often struggle with nuanced interpretation of rare variants and their clinical impact.

Purpose of the Study:

• To develop and validate machine learning (ML) models for accurate variant penetrance estimation.
• To evaluate the performance of ML-derived penetrance across different variant classes and disease-predisposition genes.
• To compare ML-based penetrance estimation with conventional case-versus-control approaches.

Main Methods:

• Construction of ML models using electronic health records from 1,347,298 participants.
• Application of ML models to an independent cohort with linked exome data.
• Evaluation of ML penetrance for 1648 rare variants in 31 autosomal dominant genes, correlating with clinical outcomes and functional data.

Main Results:

• ML penetrance estimation was variable but highest for pathogenic and loss-of-function variants.
• ML penetrance demonstrated association with clinical outcomes and functional data.
• ML approach provided refined quantitative estimates, outperforming conventional methods in interpreting variants of uncertain significance and delineating clinical trajectories.

Conclusions:

• Machine learning offers a scalable and accurate method for quantifying variant disease risk.
• Deep phenotyping combined with ML enhances the interpretation of genetic variants.
• This approach advances precision medicine by improving the understanding of genotype-phenotype relationships.