Variant-Specific Mendelian Risk Prediction Model

Area of Science:

• Genetics and Genomics
• Cancer Epidemiology
• Biostatistics

Background:

• Pathogenic sequence variants (PSVs) increase cancer risk, necessitating accurate prediction models.
• Existing Mendelian risk models often assume uniform gene-level penetrance, which may not reflect reality for genes like BRCA1/2.
• Cancer risk can vary significantly between different PSVs within the same gene.

Purpose of the Study:

• To extend Mendelian risk prediction models by incorporating PSV-specific penetrances.
• To evaluate the performance and clinical utility of the proposed Fam3PRO-variant model.
• To assess the impact of underreporting in family history data on model accuracy.

Main Methods:

• Developed the Fam3PRO-variant model, an extension of Fam3PRO, incorporating PSV-specific penetrances for BRCA1/2 variants.
• Classified BRCA1/2 PSVs into cancer-specific risk regions: breast cancer clustering region (BCCR), ovarian cancer clustering region (OCCR), and 'other'.
• Conducted simulations and evaluated the model on two large cohorts (CGN and CCGCRN), assessing calibration, discrimination, accuracy, PPV, NPV, sensitivity, and specificity.

Main Results:

• The Fam3PRO-variant model demonstrated high calibration, discrimination, and accuracy in predicting region-specific BRCA1/2 carrier status.
• The model remained robust against underreporting in family history data, providing more accurate region-specific predictions than gene-level models.
• Clinical utility assessment showed high specificity and NPV at the region-specific level, comparable to the existing gene-level model.

Conclusions:

• Mendelian risk prediction models can be effectively enhanced with PSV-specific penetrances, even with data limitations like underreporting.
• The Fam3PRO-variant model provides more precise region-specific PSV carrier probabilities, improving cancer risk prediction and prevention strategies.
• This approach offers a valuable tool for personalized cancer risk assessment and management.