Simplifying complex QSAR's (quantitative structure-activity relationships) in toxicity studies with multivariate statistics
Gerald J. Niemi, J. M. McKim · OSTI OAI (U.S. Department of Energy Office of Scientific and Technical Information) · 1988
During the past several decades many quantitative structure-activity relationships (QSAR's) have been derived from relatively small data sets of chemicals in a homologous series and selected empirical observations. An alternative approach is to analyze large data sets consisting of heterogeneous groups of chemicals and to explore QSAR's among these chemicals for generalized patterns of chemical behavior. The use of exploratory multivariate statistical techniques for simplifying complex QSAR problems is demonstrated through the use of research data on biodegradation and mode of toxic action. In these examples, a large number of explanatory variables were examined to explore which variables might best explain whether a chemical biodegrades or whether a toxic response by an organism can be used to identify a mode of toxic action. In both cases, the procedures reduced the number of potential explanatory variables and generated hypotheses about biodegradation and mode of toxic action for future research without explicitly testing an existing hypothesis.