Case study
Discovery of CDK2 inhibitors
All case studies


CDK2 is an attractive cancer drug target because its abnormal activation can promote hyper-proliferation and confer resistance to FDA-approved CDK4/6 inhibitors. First-generation inhibitors targeting CDK2 suffered from poor tolerability in the clinic likely due to off-target activities. At least five CDK2 inhibitors are currently investigated in clinical trials: PF-07104091 from Pfizer, BLU-222 from Blueprint Medicines, INX-315 from Incyclix Bio, INCB123667 from Incyte, and AZD8421 from AstraZeneca. The clear therapeutic opportunity CDK2 brings to oncology has led Pharma companies to a race for gaining first-in-class inhibitor approval.

Results and Discussion
Our approach to discovering new CDK2 inhibitors combined physics-based computer simulations, machine learning (ML), and artificial intelligence methods. Initially, we used AceDock, a scaffold docking algorithm utilizing pharmacophoric rescoring, together with a fast ML method for predicting binding affinities, and Molecular Dynamics (MD) simulations to assess the stability of the ligands. Starting from an 18 million library of fragment-size molecules, our discovery protocol selected 10 compounds to be experimentally tested after 3 working weeks. Two hits from different chemical series were discovered leading to a 20% hit rate.

Then, we used an Active Learning approach with our QuantumBind platform to expand the number of hits in a chemical series. Combining the state-of-the-art Relative Binding Free Energy (RBFE) calculations with ML model predictions allowed a quick assessment of 1K analog molecules. We prioritized the best 12 compounds for testing in only 4 weeks and found a remarkable hit rate of 92%.  


Overall, 13 hits from two chemical series were discovered after two monthly rounds by testing only 22 compounds. The hit molecules are fragment-sized with high ligand efficiency and placed in a strong IP position. The excellent success rate and timings of our hit discovery protocol demonstrate the high performance of our methods and the efficiency of our execution.