Estimation of Acid Dissociation Constants Using Graph Kernels

The biopharmaceutical profile of a compound depends directly on the dissociation constants of its acidic and basic groups. We estimate these constants using kernel ridge regression and graph kernels. The performance of our approach is similar to that of a semi-empirical model (Tehan et al, QSAR & Comb. Sci. 21(5): 457-472, 473-485) based on frontier electron theory, but uses only the annotated structure graph. In particular, no structure optimization is necessary. We discuss advantages and shortcomings of our approach.