Paper published in Molecular Pharmaceutics
“The use of artificial intelligence is going to really be a significant advance on our ability to select compounds.”
Nigel Greene, Director of Data Science & AI, AstraZeneca
Key outcomes
- Alchemite™ successfully combined descriptors, in silico, and in vitro data to predict PK parameters and curves.
- Results were comparable to the best results in the literature and a variety of in-house AZ methods.
- Direct prediction of i.v. curves was particularly accurate.
- This work can reduce time, cost, and number of animal studies in late-stage drug discovery.
Summary
Pharmacokinetics (PK) describes how the body affects a drug after administration. The concentration-time profile of a compound reflects its exposure, duration of action, safety margin and other critical factors affecting the success of a potential drug.
Accurate predictions of PK would enable better decisions regarding the selection of compounds for in vivo studies, reducing the number of experiments required, the associated costs, and the number of animal studies needed. But, this is particularly challenging because in vivo PK is influenced by many biological mechanisms.
This problem has been the focus of a project involving Optibrium and AstraZeneca, in which the Alchemite™ method was successfully applied to the prediction of PK parameters, based on compound structure and sparse in vitro data. The work has now been published in Molecular Pharmaceutics.
Publication details
Published in: Molecular Pharmaceutics 19, 5, 1488-1504 (2022)
Title: Prediction of In Vivo Pharmacokinetic Parameters and Time–Exposure Curves in Rats Using Machine Learning from the Chemical Structure
Authors: Olga Obrezanova, Anton Martinsson, Tom Whitehead, Samar Mahmoud, Andreas Bender, Filip Miljković, Piotr Grabowski, Ben Irwin, Ioana Oprisiu, Gareth Conduit, Matthew Segall, Graham F. Smith, Beth Williamson, Susanne Winiwarter, and Nigel Greene (AstraZeneca, Optibrium, Intellegens).
