Modeling ADME/Tox for Drug Discovery in the Age of Data

Absorption of drug candidates and their distribution and fate within the body are of paramount importance in determining their pharmacological potency and potential toxicities. These properties, collectively known as ADME, in addition to their toxicological risk, need to be assessed early in the drug discovery process to filter out compounds with undesired properties. ADME/tox properties are commonly estimated using in vitro and in silico methods, with in vitro methods also requiring special modeling to extrapolate to in vivo situations. Computational modeling within the field of ADME/tox has grown in popularity, ease of reach, breadth, and capacity, as well as the accuracy of its predictions. This is due to an increase in the amount of biomedical data available for modeling, as well as the advent of more sophisticated computer algorithms for molecular representation and machine learning. Here, we discuss various aspects of ADME/tox and related parameters, the available ADME/tox datasets and data considerations, the molecular features that are most relevant to ADME/tox modeling, and machine learning applications, along with the current status and future outlook within the discipline.