During the drug design process, usually main focus is put on the prevision of the adequate activity of a compound towards considered set of targets. However, it is insufficient, as despite possessing desired affinity profile, a compound might not be further considered in the drug discovery pipeline, due to unfavorable ADMET properties.

In the study, we develop a platform for comprehensive evaluation of a compound in terms of ADMET features. We consider solubility, metabolic stability, biological membranes permeability, hERG channels blocking, and mutagenicity. For all evaluated features, ligand-based models were developed (with the use of machine learning algorithm: Support Vector Machines, and key-based fingerprints for compound representation). In addition, metabolic stability, and hERG channels blocking can also be evaluated in the structure-based mode. It involves docking to the respective proteins (8 CYP subtypes in the case of metabolic stability), representation of obtained ligand-protein complexes via the Structural Interaction Fingerprints and their automatic evaluation with the use of machine learning methods.

All tools developed within the project will finally be incorporated into the on-line platform enabling their wide use by scientific community.

Acknowledgments

The study was supported by the grant OPUS 2018/31/B/NZ2/00165 financed by the National Science Centre, Poland (www.ncn.gov.pl)