Probabilistic neural network modeling of the toxicity of chemicals to Tetrahymena pyriformis with molecular fragment descriptors.

We present the results of an investigation into the use of a probabilistic neural network (PNN) based methodology to model the 48-60-h ICG50 (inhibitory concentration for population growth) sublethal toxicity to the ciliate Tetrahymena pyriformis. The information fed into the neural network is solely based on simple molecular descriptors as can be derived from the chemical structure. In contrast to most other toxicological models, the octanol/water partition coefficient is not used as an input parameter and no rules of thumb, or other substance selection-criteria, are involved. The model was trained on a 1,000 substances data set and validated using an 84 substances external test set. The associated analysis of errors confirms the excellent recognitive and predictive capabilities of the model.