Abstract
The ever-increasing costs of in vitro and in vivo testing are compelling scientists to increasingly rely on computational models for predictive characterisation at early stages of drug discovery and development. The complexity of this stage requires high-throughput screening methods that can rapidly generate comprehensive information about new chemical compounds. This review explores innovative approaches assessing pharmacokinetic and pharmacodynamic properties of new chemical entities, with a focus on integrating machine learning as a transformative analytical tool. Machine learning algorithms are highlighted for their capability to train sufficient predictors combining biomimetic chromatography data (a high-throughput alternative for several physicochemical assays) with molecular features and/or molecular fingerprints obtained in silico and in vivo data of known compounds to allow efficient prediction of in vivo data for new chemical entities. By synthesising recent methodological advancements and giving useful practical approaches, the review provides insights into computational strategies that can significantly accelerate compound library screening and drug development processes.