Chapter 23 - QSAR models for predicting in vivo reproductive toxicity

Abstract

Reproductive toxicity testing is vital in product safety evaluation and chemical risk assessment and includes multiple complex toxicity endpoints. In vivo animal studies are the main method for reproductive toxicity testing. However, an animal study for testing reproductive toxicity may utilize multiple generations of animals, require several years, and is expensive and time-consuming. Additionally, the ethical concerns of animal use and the difficulty to extrapolate the results from animals to humans demand nonanimal alternative approaches. Therefore, machine learning has been explored as an alternative approach for reproductive toxicity prediction. Quantitative structure–activity relationship (QSAR) models for reproductive toxicity prediction have been established based on the limited reproductive toxicity data from the European Chemicals Agency-Classification and Labelling Inventory, toxicity reference database, Procter and Gamble research data, and the Leadscope database. Tools of QSAR models for reproductive toxicity prediction have been developed and evaluated using pesticides. This chapter introduces the QSAR models that have been developed for predicting reproductive toxicity and reviews the application of the available tools of QSAR models for reproductive toxicity prediction.