Machine Learning-Based Virtual Screening and QSAR Analysis of ESR1 Inhibitors for Breast Cancer Treatment

Abstract

Estrogen receptor alpha (ESR1) is a critical therapeutic target in hormone receptor positive breast cancer, however, endocrine resistance remains a challenge. The aim of this study is to identify potential inhibitors of the oestrogen receptor (ESR1) from bioactivity data from ChEMBL using an integrated computational workflow. A collection of 3130 compounds that have been reported with IC50 values were subjected to pIC50 transformation, PubChem fingerprint encoding, Random Forest QSAR modeling, Lipinski screening, ADMET profiling, molecular docking, and molecular dynamics simulation. The Random Forest model showed acceptable predictive performance, with a test-set R² of 0.738 and cross-validated Q² values of 0.625 and 0.638. After filtering, 115 candidates were prioritized for docking against ESR1. Docking analysis identified CHEMBL2403358 as the top candidate, exhibiting a binding affinity of 10.0 kcal/mol, which surpassed that of the reference ligand, estradiol. This molecule established key interactions with residues GLU353, ASP351, LEU387, LEU391, and PHE404. MD simulations corroborated the stability of the ESR1–CHEMBL2403358 complex throughout the trajectory. This integrated in silico workflow designates CHEMBL2403358 as a promising ESR1 inhibitor scaffold, warranting further experimental validation for the development of novel therapeutic strategies against endocrine-resistant breast cancer.