Christopher Cassa
Assistant Professor of Medicine, Brigham and Women’s Hospital, Harvard Medical School
Assessing clinical risk using polygenic risk and monogenic variants
Clinical risk assessment remains challenging, even for patients with variants in well-known genes such as BRCA1 or LDLR. In these genes, many variants are so rare that it is challenging to provide a reliable assessment of risk, making them difficult to translate into clinical care. Previous studies have demonstrated the utility of polygenic risk scores (PRS) in clinical risk assessment and in the variable penetrance or expressivity of monogenic variants. Here, we develop a model which integrates individual-level clinical risk factors (e.g. PRS, sex, family hisory) with variant-level predictive features (e.g. consequence, computational predictions, frequency, context). We fit a Cox regression with patient outcomes for 49,738 individuals with exome sequence data (UK Biobank). The resulting risk estimates are highly concordant with individual clinical outcomes and can significantly distinguish VUSs that carry elevated clinical risk from those with population-level disease risk (log-rank p=3×10-5). For patients who carry previously identified pathogenic variants (ClinVar), inclusion of individual-level characteristics separates patients truly at higher risk (𝛘2 p=0.001) versus those with no significant elevation in clinical risk (𝛘2 p=0.15). Such risk assessments may be used to optimize clinical surveillance strategies and intervention.
