Chemoinformatics Study of Dimedone‐Derived Compounds as Inhibitors of Human Colon Cancer HT‐29

Abstract

Colorectal cancer (CRC) remains a major health burden worldwide, motivating the search for safe and effective small‐molecule therapeutics. Here, we report an integrated chemoinformatics study on 34 dimedone‐derived compounds active against HT‐29 colon cancer cells. RDKit was used to compute 174 descriptors, and a wrapper selection with Hyperopt (TPE) identified a four‐descriptor multiple linear regression model (PEOE_VSA11, PEOE_VSA7, VSA_EState10, fr_NH1) that explained the biological variance well ( R 2  = 0.759; adjusted R 2  = 0.705; F  = 14.16, p  = 2.16 × 10 − 5 ) with minimal multicollinearity (VIF ≈ 1). Positively weighted van der Waals surface and electrotopological descriptors were the dominant drivers of potency, whereas fr_NH1 contributed slightly negatively. Guided by these trends, we designed derivatives Z7 and Z9, predicted to improve potency while maintaining reasonable developability. Molecular docking against the receptor tyrosine kinase c‐MET showed strong binding for the reference M34 (−9.3 kcal/mol) and favorable binding for Z9 (−8.0 kcal/mol) through complementary hydrophobic and electrostatic contacts. One hundred nanosecond GROMACS simulations indicated that both cMET_M34 and cMET_Z9 complexes are dynamically stable; Z9 promotes sustained one‐to‐two hydrogen bonds and a gradual reduction in solvent accessibility, consistent with deeper burial. A reinforced convergence analysis confirms that the 100 ns trajectories are well equilibrated.