3D Electron Cloud Descriptors for Enhanced QSAR Modeling of Anti-Colorectal Cancer Compounds

To address the limitations of conventional Quantitative Structure-Activity Relationship (QSAR) descriptors in capturing the electronic and spatial complexity of molecular structures, we developed a high-dimensional QSAR modeling framework based on three-dimensional electron density features.Electron densities were computed using density functional theory (DFT), transformed into 3D point clouds, and encoded into a multi-scale descriptor set incorporating radial distribution functions, spherical harmonic expansions, point feature histograms, and persistent homology. This design captures molecular characteristics across statistical, geometric, and topological levels. Across multiple machine learning models, the proposed descriptors consistently enhanced predictive performance; for instance, Area Under the Curve (AUC) improved from 0.88 to 0.96 with Light Gradient Boosting Machine (LightGBM) (effect size: 0.075-0.079).Feature attribution analysis identified local geometric descriptors and intensity-based electronic features as primary contributors. The integration of these descriptors with traditional 1D/2D features further improved accuracy, demonstrating their strong complementarity. Model robustness and reliability were validated through DeLong and permutation tests, calibration curve assessments, and applicability domain analysis.This study establishes an electron density–driven paradigm for molecular representation, offering a promising computational avenue for precision-oriented anticancer drug discovery.