Prediction of inherited genomic susceptibility to 20 common cancer types by a supervised machine-learning method

Significance Prevention and early diagnosis of cancer are the most effective ways of avoiding psychological, physical, and financial suffering from cancer. We present a machine-learning method for statistically predicting individuals’ inherited susceptibility (and environmental/lifestyle factors by inference) for acquiring the most likely type among a panel of 20 major common cancer types plus 1 “healthy” type. The results show that, depending on the type, about 33–88% of a cancer cohort have acquired its cancer type primarily due to inherited genomic susceptibility factors and that the rest are primarily due to environmental/lifestyle factors. These personal genomic susceptibilities with associated probabilities may provide practical information for individuals, heath professionals, and health policymakers related to prevention and/or early intervention of cancer.