Helping the FDA utilize the most current scientific knowledge in its regulatory processes is central to the NCTR³s mission. Development of Knowledge Bases that have predictive values to support regulatory decision-making, and that help identify data and research gaps is a key strategic goal. The first such computer-based system under development is for prediction of estrogenicity. A 3D-QSAR model using Comparative Molecular Field Analysis (CoMFA) was built from data for Estrogen Receptor Binding Affinity (RBA) in calf uterine cytosol. This model yielded a statistically robust cross-validation for prediction in a test set for RBA in human MCF-7 cell cytosol. The method of Principal Components Analysis and neural networks supervised learning using quantum chemical descriptors has produced models with utility for simple classification. Estrogenic compounds are ubiquitous in nature, and are of increasing concern for the FDA because of their potential for disruption of the endocrine system in general. Human and animal exposure of potential concern to FDA includes anabolic agents in livestock feed, soy phytoestrogens, pesticide and herbicide residues, cosmetic additives, estrogenic plasticizers in devices, and numerous pharmaceuticals.