Probabilistic Predictive Model of the Human Liver Microsomal Metabolism Regioselectivity

Moreover, in silico predictions of the most likely metabolism sites in a molecule could facilitate the
analysis of spectroscopic data and thus ease the experimental identification of metabolites.

In this work, we present QSAR models for the prediction of metabolism regioselectivity. They provide the probability to be metabolized in human liver microsomes for every atom of the molecule and are based on a novel GALAS (Global, Adjusted Locally According to Similarity) methodology – an approach enabling the evaluation of Model Applicability Domain via the calculation of the prediction Reliability Index (RI).