Type 2 Diabetes Risk Prediction Incorporating Family History Revealing a Substantial Fraction of Missing Heritability

Despite many successes of genome-wide association (GWA) studies, known susceptibility variants identified by GWAS have the modest effect sizes and we met noticeable skepticism about the risk prediction model building with large-scale genetic data. However, in contrast with genetic variants, family history of diseases has been largely accepted as an important risk factor in clinical diagnosis and risk prediction though; complicated structures of family history of diseases have limited their application to clinical use. Here, we develop a new method which enables the incorporation of general family history of diseases with the liability threshold model and a new analysis strategy for risk prediction with penalized regression incorporating large-scale genetic variants and clinical risk factors. An application of our model to type 2 diabetes (T2D) patients in Korean population (1846 cases out of 3692 subjects) demonstrates that SNPs accounts for 28.6% of T2Ds variability and incorporation of family history leads to additional improvement of 5.9%. Our result illustrates that family history of diseases can have an invaluable information for disease prediction and may bridge the gap originated from missing heritability.