Sohvi Luukkonen Abstract

DrugEx: Deep Learning for de novo Drug Design – A Case for A2B Selective Ligands

Sohvi Luukkonen1, Martin Sicho1,2, Helle W. Maagdenberg1,  Linde Schoenmaker1,  Olivier J.M. Béquignon1,  Jerre Madern1,  Daan van der Es1 and Gerard van Westen1

1 Leiden Academic Centre of Drug Research, Leiden University, 55 Einsteinweg, 2333 CC Leiden, The Netherlands
2 bCZ-OPENSCREEN: National Infrastructure for Chemical Biology, Department of Informatics and Chemistr, Faculty of Chemical Technology, University of Chemistry and Technology Prague, Technick ́a 5, 166 28, Prague, Czech Republic


The discovery of novel molecules with desirable properties is a classic challenge in medicinal chemistry. With the recent advancements in machine learning, there has been a surge of de novo drug design tools. However, few resources exist that are both user-friendly as well as easily customizable. Furthermore, out of the plethora of recently developed de novo drug design tools, very few have been experimentally validated. Our group has developed a versatile open-source software package DrugEx for multi-objective reinforcement learning [1]. The package contains multiple generator architectures and a variety of scoring tools and multi-objective optimisation methods and is a continuation of the original incremental work of Liu et al.’s DrugEx [2–4].

The activation of the A2B, an isoform of adenosine receptors (ARs), has been linked to hallmarks of cancer. Therefore, antagonizing the A2B could be a promising strategy to tackle some cancers. However, therapeutically available antagonists are currently non-selective for A2B, resulting in unwanted off-target effects. Ligands with a tri-substituted pyrazine scaffold have been shown to have selectivity for A2B over the other ARs [5].

In this work, we have used DrugEx to generate A2B selective ligands from a pyrazine scaffold input, synthesized a selection of them and will measure their activity against all ARs. We not only provide nice software but also actually follow up with prospective validation in the wet lab in a real project.

(1) Sicho, M. et al. DrugEx: Deep Learning Models and Tools for Exploration of Drug-like Chemical Space. 2023, in preparation.
(2) Liu, X. et al. An exploration strategy improves the diversity of de novo ligands using deep
reinforcement learning: a case for the adenosine A2A receptor. Journal of Cheminformatics
2019, 11, 35.
(3) Liu, X. et al. DrugEx v2: de novo design of drug molecules by Pareto-based multi-objective reinforcement learning in polypharmacology. Journal of Cheminformatics 2021, 13, 85.
(4) Liu, X. et al. DrugEx v3: Scaffold-Constrained Drug Design with Graph Transformer-based
Reinforcement Learning. 2021, DOI: 10.26434/chemrxiv-2021-px6kz.
(5) Eastwood, P. et al. Discovery of LAS101057: A Potent, Selective, and Orally Efficacious A2B Adenosine Receptor Antagonist. ACS Medicinal Chemistry Letters 2011, 2, Publisher: American Chemical Society, 213–218.