Combined machine learning models, docking analysis, ADMET studies and molecular dynamics simulations for the design of novel FAK inhibitors against glioblastoma.

Gliomas, particularly glioblastoma (GBM), are highly aggressive brain tumors with poor prognosis and high recurrence rates. This underscores the urgent need for novel therapeutic approaches. One promising target is Focal adhesion kinase (FAK), a key regulator of tumor progression currently in clinical trials for glioma treatment. Drug development, however, is both challenging and costly, necessitating efficient strategies. Computer-Aided Drug Design (CADD), especially when combined with machine learning (ML), streamlines the processes of virtual screening and optimization, significantly enhancing the efficiency and accuracy of drug discovery. Our study integrates ML, docking analysis, ADMET (absorption, distribution, metabolism, elimination, and toxicity) studies to identify novel FAK inhibitors specific to GBM. Predictive models showed strong performance, with an R(2) of 0.892, MAE of 0.331, and RMSE of 0.467 using protein-level IC(50) data in combined CDK, CDK extended fingerprints, and substructure fingerprint counts derived from 1280 FAK inhibitors. Another model, based on IC(50) data from 2608 compounds tested on U87-MG cells, achieved an R(2) of 0.789, MAE of 0.395, and RMSE of 0.536. Using these models, we efficiently identified 275 potentially active compounds out of 5107 candidates. Subsequent ADMET analysis narrowed this down to 16 potential FAK inhibitors that meet the established drug-likeness criteria. Moreover, molecular dynamics (MD) simulations validated the stable binding interactions between the selected compounds and the FAK protein. This study highlights the effectiveness of combining ML, docking analysis, and ADMET studies to rapidly identify potential FAK inhibitors from large databases, providing valuable insights for the systematic design of FAK inhibitors.