ML-Based Atom-Centered 3D-QSAR, Docking, ADME and DFT Studies of Amino-Pyrimidine Derivatives as FGFR4 Anticancer Inhibitors

In the present study, amino-pyrimidine–based compounds were computationally investigated as potential FGFR4 inhibitors using Schrödinger molecular modelling tools. An atom-based machine learning-driven 3D-QSAR model was developed and validated, demonstrating excellent statistical robustness (training set: R² = 0.996; test set: Q² = 0.774; F = 1211.5; RMSE = 0.68; SD = 0.106). Molecular docking studies revealed favourable binding interactions with the FGFR4 active site, highlighting compound S1 as the most promising candidate with a high docking score (−8.660). Density functional theory (DFT) analysis showed that S1 possesses a low-lying HOMO, a small energy gap (ΔE = 3.37 eV), and enhanced electronic stability, supporting its potential biological activity. In-silico ADME predictions further indicated favourable pharmacokinetic properties, including high oral absorption (80–90%), compliance with Lipinski’s rule of five, moderate metabolic stability, and acceptable blood–brain barrier permeability. Overall, the integrated ML-based 3D-QSAR, docking, ADME, and DFT analyses identify S1 as a promising FGFR4 inhibitor. In summary, the combined computational results indicate that S1 represents a promising virtual lead candidate for FGFR4 inhibition. However, all findings are predictive in nature and are intended to guide future experimental efforts. Experimental validation, including biochemical and cellular assays, is required to confirm the actual biological activity and therapeutic potential of the proposed compound.