Integrating machine learning with in silico studies and Quantum Chemistry: Exploring novel compounds through multiscale screening targeting the CDK2 enzyme.

Cyclin-dependent kinase 2 (CDK2) modulates the progression of the cell cycle, and its dysregulation results in unchecked cellular proliferation, establishing it as a pivotal target in oncological therapies. We implemented a comprehensive screening pipeline to identify potential novel inhibitors for the CDK2 enzyme by integrating advanced machine learning classification methods. The random forest (RF) method shows better performance based on the statistical metrics assessment. This RF model was used to filter a large coconut dataset comprising 477,975 molecules to identify potential candidates. This initial screening process identified 327 candidate molecules. The subsequent application of PAINS (Pan-Assay Interference Structures) filtration refined this list to 309 molecules, which were then selected for molecular docking analysis. Based on the docking score, the top 40 potential candidates from molecular docking analysis were chosen for pharmacokinetics (PK) and pharmacodynamics (PD) studies (ADMET). Three molecules that satisfy the PK/PD criteria were selected for DFT and molecular dynamics simulation studies. The finalized three molecules displayed conserved interactions with the residues Lys33 and Asp145, crucial for enzyme inhibition. Moreover, Molecule 2 possessed an extended fused heterocyclic system, which may enhance its inhibitory potential. The simulation studies indicate that these compounds showed stable behavior within the binding pocket of the CDK2 enzyme. Also, we have developed an open-access online tool named "pCDK2i_v1.0" to help the scientific community efficiently screen the potential CDK2 inhibitors. This work demonstrates the importance of integrating machine learning in drug design to discover novel anti-cancer inhibitors of the CDK2 enzyme. The pCDK2i_v1.0 tool for screening and predicting the CDK2 activity as active (1), and inactive (0) is available at https://github.com/Amincheminfom/pCDK2i_v1.