Binding affinity between chemicals and the estrogen receptor (ER) serves as an indicator of the potential to cause endocrine disruption through this receptor-mediated endocrine pathway. Estimating ER-binding affinity is, therefore, one strategic approach to reducing the costs of screening chemicals for potential risks of endocrine disruption. While measuring ER binding with in vitro assays may be the first choice in prioritizing chemicals for additional in vitro or in vivo estrogenicity testing, the time and costs associated with screening thousands of chemicals is prohibitive. Recent advances in 3D modeling of the reactivity of flexible structures make in silico methods for estimating ER binding possible. One technique, the common reactivity pattern (COREPA) approach, was applied to development of reactivity patterns for ER relative binding affinity based on global nucleophilicity, interatomic distances between nucleophilic sites, and local electron donor capability of the nucleophilic sites. The reactivity patterns provided descriptor profiles for order-of-magnitude RBA ranges of training set chemicals. An exploratory expert system was subsequently developed to predict RBA and rank chemicals with respect to potential estrogenicity. A strategy is presented for extending initial exploratory 3D QSAR models beyond current training sets to increase applicability to more diverse structures in large chemical inventories.