Abstract 1151 Machine Learning for understanding enzyme function and effects of cancer-associated mutations

Our Machine Learning methodology, Partial Order Optimum Likelihood (POOL), is used to predict biochemically active amino acids in the 3D structures of proteins. Computed electrostatic and chemical properties of individual amino acids serve as input features. Our most recent applications of POOL are described. POOL analysis uncovers the types of interactions that enable protein structures to transform the weakly acidic or basic side chains of amino acids into strong acids, strong bases, or nucleophiles in the active site. This analysis can also predict whether a mutation is likely to alter biochemical function. The μ4 value, a metric obtained from POOL and THEMATICS that serves as a measure of the degree of coupling between one ionizable amino acid and its neighbors, is used to identify which protein mutations are likely to have significant impact on the biochemical activity. For examples of cancer-associated variants of human DNA polymerase kappa (pol kappa), THEMATICS and POOL are used to predict which ones have altered biochemical activity. Kinetic assays on undamaged and damaged DNA confirm predictions of reduced activity for the pol kappa variants R48I, H105Y, G147D, G154E, V177L, R298C, E362V, and R470C relative to wild type. The pol kappa variants T102A, H142Y, R175Q, E210K, Y221C, N330D, N338S, K353T, and L383F were predicted to have similar catalytic efficiency to WT pol kappa; these predictions were confirmed by experiment in vitro. This raises intriguing questions about the mechanisms for disease association in vivo. This work is supported by NSF MCB-2147498.