Functional impact of missense variants in BRCA1 predicted by supervised learning

Area of Science:

• Genetics
• Bioinformatics
• Computational Biology

Background:

• Many individuals tested for inherited cancer susceptibility have BRCA1 variants of unknown clinical significance (UCVs).
• Biochemical assays for UCVs are time-consuming and labor-intensive.
• Computational methods can complement functional tests by predicting UCV impact.

Purpose of the Study:

• To develop and evaluate a supervised learning approach for classifying BRCA1 UCVs.
• To complement biochemical assays and improve the interpretation of UCVs.
• To provide a tool for clinicians to better inform patients about genetic risks.

Main Methods:

• A supervised learning approach using 16 predictive features.
• Retrospective classification of 36 BRCA1 C-terminal (BRCT) domain UCVs with known biochemical assay results.
• Blind classification of 54 documented BRCA1 UCVs.
• Majority vote of three supervised learning algorithms.

Main Results:

• The computational classifiers agreed with biochemical assays for over 94% of UCVs.
• Two UCVs classified as deleterious by both methods revealed a novel binding site.
• The approach demonstrated high accuracy in classifying BRCA1 UCVs.

Conclusions:

• Supervised learning algorithms can accurately classify BRCA1 UCVs, complementing functional assays.
• This computational approach can aid in patient genetic counseling and risk assessment.
• The methodology is adaptable to other cancer susceptibility genes and large-scale genetic studies.