Network pharmacology and machine learning nominate PPARG as a candidates target of Aegle marmelos (L.) Correa phytochemicals in breast cancer

Abstract

Breast cancer remains a leading cause of cancer death in women, and toxicity, tumor heterogeneity and acquired resistance sustain the search for multi-targeted agents of natural origin. Aegle marmelos (L.) Correa is a traditional medicinal species whose molecular targets in breast cancer have not been defined systematically. Filtering 138 reported phytochemicals on drug-likeness, oral bioavailability and predicted toxicity left six leads, whose 252 predicted targets yielded ten hub genes by protein–protein interaction degree. Three classifiers trained on GSE42568 (104 tumor, 17 normal) under nested tenfold cross-validation reached pooled areas under the curve of 0.939–0.974, and their marker sets intersected in PPARG and ERBB2; a composite score adding differential expression, network degree and out-of-fold SHAP importance ranked BCL2 a clear second, with EP300, EGFR and ERBB2 clustered within 0.02 of one another for third, too close to call. All compounds were docked with six targets; marmeline was the strongest lead at four of the six receptors. Molecular dynamics and Hartree–Fock-level quantum-chemical characterization, completed for marmeline at PPARG against its own co-crystallized reference ligand, rosiglitazone, found both complexes stable over 200 ns (backbone RMSD 2.4–2.6 Å) with MM/PBSA binding free energies of −26.5 and −29.6 kcal mol−1, and confirmed marmeline's optimized geometry as a genuine minimum (zero imaginary frequencies; HOMO–LUMO gap 9.97 eV). Together, these findings identify PPARG as the best-supported target by almost every measure and marmeline as a structurally supported lead for experimental validation.