Combined machine learning models, docking analysis, molecular dynamics and experimental validation for the rapid design of novel FLT3 inhibitors against AML

Abstract Acute myeloid leukemia (AML) is a malignant clonal disorder driven by the excessive proliferation of immature myeloid cells in the bone marrow and blood, often linked to Fms-like tyrosine kinase 3 (FLT3) mutations, which occur in about one-third of AML patients. While FLT3 inhibitors such as Midostaurin, Quizartinib, and Gilteritinib have demonstrated clinical efficacy, their therapeutic potential is often limited by drug resistance and adverse reactions. Therefore, the development of novel FLT3 inhibitors is critical for improving AML treatment outcomes. In this study, we employed a multi-faceted computer-aided drug design (CADD) approach, integrating machine learning (ML), molecular docking, and molecular dynamics simulations, to accelerate the discovery of new FLT3 inhibitors. A machine learning-based FLT3 classification model achieved an accuracy of 0.958, while an MV4-11 cell activity prediction model demonstrated strong predictive performance with an R2 of 0.846, MAE of 0.368, and RMSE of 0.492. Virtual screening of 7,280 compounds from the ChemDiv database led to the identification of 68 potential FLT3 inhibitors, with molecular dynamics simulations confirming their stable binding to the FLT3 protein. Experimental validation of four selected compounds showed promising activity in MV4-11 cellular assays, demonstrating the reliability of this integrated CADD approach. These results underscore the potential of a CADD-driven approach, enhanced by ML, to rapidly design new FLT3 inhibitors for AML treatment.