Accelerated discovery of novel RORgammaT modulators using an AI-driven platform integrating generative chemistry and mechanistic PK simulation

The application of artificial intelligence (AI) in oncology drug discovery offers the potential to significantly accelerate and de-risk the identification of novel therapeutic agents. In this study, we present the successful application of an AI-driven drug design (AIDD) platform to discover new classes of RORyT ligands, both agonists and inverse agonists. RORyT, a nuclear receptor central to Th17 cell differentiation and IL-17 signaling, has emerged as a promising target for auto-immune diseases as well as cancer immunology, and both classes of ligands have therapeutic potential. Our AIDD platform integrates quantitative structure-activity relationship (QSAR) modeling, ADMET property predictions, high-throughput mechanistic PK simulations, 3D volumetric/pharmacophore similarity scoring, and synthetic accessibility assessments at the point of initial compound design. Importantly, compound prioritization is guided by a multi-criteria decision analysis multi-parameter optimization (MPO) algorithm, which incorporates these features to balance potency, ADMET/PK, and chemical tractability to optimize candidate selection. This approach enables systematic decision-making across multiple parameters, reducing reliance on trial-and-error screening. In our initial Design/Make/Test cycle, 27 novel compounds were synthesized and tested. Remarkably, 70% demonstrated >25% inhibition of RORyT activity in cell-based assays. The top candidate had an IC₅₀ of 1.51 uM and favorable in vitro ADME, with high concordance between predicted and measured values. Based on these results, a second set of inverse agonists was designed. 14/19 (74%) were active, with 4 having improved potency. In vitro and in vivo ADMET and PK testing is ongoing. In parallel work, we identified a novel class of RORyT agonists, which could find application in cancer immunotherapy; further development of this class is ongoing as well. This work demonstrates how a platform integrating generative chemistry with mechanistic PK simulations and MPO can be used to accelerate the development of viable drug candidates, in this case, against an important clinical target, RORyT