Computational and Medicinal Chemistry by the Lake 2026

Towards a Comprehensive De Novo Design Approach: The Discovery of p38:MK2 Molecular Glues

Speaker:
Márton Vass, Senior Principal Scientist, Schrödinger

Abstract:
Modern drug discovery must still contend with many challenges including a fast-moving competitive landscape, high development costs, and low project success rates. Schrödigner’s computational platform addresses these challenges, leveraging physics-based methods and AI/ML workflows to accelerate DMTA cycles. Schrödinger’s de novo design approach utilizes AutoDesigner [1,2] for ultra-large scale chemical space exploration, generating and prioritizing billions of novel structures. This is combined with rigorous free energy calculations using FEP+ [3] in an iterative fashion (Active Learning FEP+) to rapidly score libraries and prioritize design ideas with optimal potency. Here, we show the successful application of this integrated workflow in the discovery of p38:MK2 molecular glues, enabling systematic scaffold and R-group exploration and leading to potent and selective hits that showed pronounced TNFα reduction in in vivo models. The integrated use of Generative ML is also effectively shown on the same project where, combined with Autodesigner and the predictive power of Active Learning FEP+, rapidly leads to the generation of novel molecules that are optimized not only for potency but also for other project endpoints. To promote the design of molecules that can be easily synthesized, we describe a novel retrosynthesis method, RetroSynth, that deconstructs complex target molecules into commercially available starting materials, prioritizing cost-efficient synthetic routes and leveraging existing project chemistry. Internal validation shows RetroSynth significantly outperforms alternative solutions in terms of accuracy and real-world drug discovery projects feasibility. Altogether, the presented tools compose a fully-integrated, comprehensive de novo design workflow that efficiently generates synthetically tractable ideas that satisfy project requirements, ultimately accelerating lead optimization.