LiADME: An Open-Source Tool for the Prediction of Pharmacokinetically Relevant Properties of Small Molecules

TALEVI A.; CARAM ROMERO F.N.; PRADA GORI D. N.; FALICO M.; ALBERCA L. N.; BELLERA C.

Montevideo

Encuentro; SETAC Latin America 15th Biennial Meeting; 2023

Early assessment of PK-related properties of small molecules is of great value to prioritize drug candidates, guide their molecular optimization and select dosing schedules during preclinical development and clinical trials. Furthermore, in silico predictions on ADME properties may be used as input for elaborate PK models. Here, we report the development of an open-source machine learning-based webapp publicly available in out institutional website as a part of LIDeB tools ( https://lideb.biol.unlp.edu.ar/?page_id=1076), capable of predicting a wide array of ADME properties, including aqueous solubility, log P, log D, tissue/blood partition coefficients, fraction bound to plasma proteins, total clearance, organ clearance, apparent volume of distribution, CYPs liability, and many others. Each predictor has been inferred and validated from a chemically diverse, representatively sampled curated dataset, and each prediction is associated with a reliability flag based on applicability domain assessment. In line with the open-source philosophy, our code can be assessed and modified for improvement, expansion, or to adjust to specific needs. We provide regular maintenance and free user support. Also noteworthy, our suite of predictors covers some frequently overlooked ADME properties, such as affinity for pharmacokinetically relevant SLC transporters and nuclear receptors associated with enzyme and transporter induction.