Dimerization-Driven Quantitative Modeling of Endocrine Disruption via Sex Hormone Receptors: Divergent Predictions for Estrogen Receptor α and Androgen Receptor

Interactions between endocrine-disrupting chemicals (EDCs) and nuclear receptors (NRs) are extensively used in virtual screening. However, most models overlook receptor dimerization. Here, we systematically investigate the dimerization dependencies of estrogen receptor α (ERα) and androgen receptor (AR) via big data mining, reporter gene assays, and molecular simulations and further construct a dimerization-driven quantitative framework for endocrine disruption prediction. Database analysis revealed that ERα predominantly functions as homodimers whereas AR primarily exists as monomers. Experiments and simulations further demonstrated that ERα forms stable dimeric complexes (agonist–ERα dimer–coactivator and antagonist–ERα dimer–none systems) while AR maintains stability as a monomer (ligand–AR monomer–coregulator system). Based on these mechanistic insights, receptor-specific quantitative models were developed and applied to predict the endocrine-disrupting potential of representative EDCs, including bisphenols and hydroxylated polybrominated diphenyl ethers. These dimerization-informed models showed robust performance for ERα dimers (R2 = 0.605–0.726), whereas AR’s monomeric configuration yielded superior predictive accuracy (R2 = 0.788–0.909). These predictive patterns match their mechanisms. Estrogenic disruption critically depends on dimerization and coactivator recruitment, while antiestrogenic activity appears to be primarily dependent on dimerization and largely independent of corepressor involvement. In contrast, AR disruption is primarily driven by ligand binding and coregulator interactions with minimal dimerization involvement. This framework provides an effective, mechanism-grounded alternative to EDC screening.