Prediction of toxicological interactions in a binary mixture by using pattern recognition techniques: Proposed approach with a developed model

A model has been suggested to predict the nature of toxicological interaction in binary mixtures. The approach uses the NLM‐HSDB computerized data bases, the ATSDR toxicologic profiles, and other literature for categorizing the nature (synergistic, antagonistic or no interaction) and degree of interaction. Multivariate modeling (pattern recognition techniques) is the statistical approach utilized to separate groups of compounds into those that interact synergistically or antagonistically with a given toxic compound. Preliminary results indicate (1) that there are sufficient data in the literature on interactions to permit such modeling and (2) that in the case of carbon tetrachloride those compounds that interact synergistically with it are more similar to each other than those that interact antagonistically with respect to a number of structural and toxicologic parameters. This suggested approach of utilizing pattern recognition tools will be quite useful for regulatory agencies in predicting toxicological interactions occurring in complex chemical mixtures in the environment.