Computational evaluation of endocrine activity of biocidal active substances

Abstract Exposure to endocrine disrupting chemicals is an important public health concern although only a few disruption have been identified so far. To speed up their identification, in silico toxicological models appear be the most appropriate, since potential of large number compounds can estimated short time. In this study three (Endocrine disruptome software, VirtualToxLab and COSMOS KNIME) used. predictions biocidal active substances made then compared with available in vitro experimental binding affinities androgen, estrogen, glucocorticoid thyroid receptors. The chosen had similar accuracies (around 60%), while differences were shown between specificity sensitivity. was balanced model. Additionally, combined prepared evaluated. As expected, majority rule approach model more accurate balanced. However, positive consensus model, that improved (?80% for all studied nuclear receptors) applicable. This reduction false especially useful search (active) compounds. On other hand, novel negative prediction receptors), giving good (inactive) excluded from further testing. results obtained by these great added value, they significantly reduce