Submission Details

Design Rationale:

In the frame of a collaboration with the Czodrowski AG and Idorsia Pharmaceuticals, the submitted compounds were designed with a Deep Generative Model (DGM) using Reinforcement Learning (RL). Using Covid Moonshot data and ChEMBL data, privileged fragments were identified by Matched Molecular Pairs and a scoring components was developed. Crystal structures (Fragalysis data) with non-covalent ligands were selected and used for a shape and 3D-Pharmacophore scoring component. The scoring components were used in RL fashion to train a DGM to generate potential SARS-COV-2 Mpro inhibitors. Molecules were then generated. The molecules were prioritized by Machine Learning for synthesizability and binding affinity. Finally, the generated molecules were further assessed by template docking using 3 selected PDB files (MPro-x12692, MPro-x11294 and Mpro-P0009) covering 3 main series (amino-pyridine, Ugi and quinolone series). 157 molecules were then selected and submitted to Covid Moonshot.