IDEAYA Biosciences, Inc. announced a research collaboration with ATTMOS as part of its efforts to build a physics-based computational small molecule discovery platform that rapidly unlocks what are currently perceived as undruggable oncology targets.

The collaboration will integrate IDEAYA's differentiated and proven capabilities in structural biology and pharmaceutical drug discovery across multiple first-in-class oncology targets with ATTMOS's capabilities in computational chemistry method development, high performance computing, and software development. The focus of the partnership will be to engineer and optimize a workflow solution for high-throughput absolute binding free energy perturbation predictions (ABFEP) of first-in-class drug candidate molecules.

This approach enables application of gold-standard physics-based statistical mechanics calculations of protein-ligand affinities at the scale required for virtual screens and represents what could become the industry's go-to standard for high-speed and high probability-of-success drug hit-finding against structurally-enabled novel biological targets.

"Current AI/ML-enabled drug discovery approaches have been largely applied to either already drugged targets or well-understood biological target classes and often fail when applied to first-in-class target opportunities. IDEAYA continues to enhance its computational drug discovery capabilities to pursue first-in-class oncology targets that are perceived as undruggable," said Michael White, Ph.D., Chief Scientific Officer, IDEAYA Biosciences. "Our partnership with ATTMOS will enable us to apply the principles of engineering to the field of drug discovery, at scale, for efficient prosecution of unprecedented oncology targets," said Paul Barsanti, Ph.D., Chief Technology Officer, IDEAYA Biosciences.

The collaboration will leverage the Amber molecular dynamics suite as the GPU-accelerated back-end free energy simulation engine. IDEAYA will train and evaluate ABFEP-based active learning cycles based on extensive ground-truth data sets derived from its successful wet-lab drug discovery campaigns against novel targets.

These models will be used to screen enormous libraries of synthetically tractable chemical space for accurate and efficient de novo discovery of small molecule ligands for new targets. The work aims to overcome the limitations of current virtual screening approaches and accelerate the discovery of novel small molecule oncology therapeutics that address unmet clinical need.