Neural network methodologies allow the modeling of nonlinear relationships. This makes them useful tools for the analysis of larger data sets of non-congeneric compounds with unknown or varying modes of action. This brief review describes recent advances and their applications to sets of several hundred to over 1 000 compounds, modeling acute toxicity data for several aquatic species, including fish, ciliate, bacteria, and non-acute toxicity data for a mammalian species endpoint, i.e. estrogen receptor binding assay data.