On the nature, evolution and future of quantitative structure-activity relationships (QSAR) in toxicology

• Published in: SAR and QSAR in environmental research Vol. 15; no. 5-6; pp. 323 - 330
• Main Author: Veith, G.D.
• Format: Journal Article
• Language: English
• Published: England Taylor & Francis Group 01.10.2004
• Subjects: Animals; Applicability domain; Drug Design; Forecasting; Models, Biological; Models, Theoretical; Nature; QSAR; Quantitative Structure-Activity Relationship; Reach; Risk Assessment; Toxicity test data; Toxicology - methods; Toxicology - trends
• ISSN: 1062-936X; 1029-046X
• DOI: 10.1080/10629360412331297380

The quantitative structure-activity relationship (QSAR) science agenda is being determined by its skeptics. Toxic substances control legislation over the past 30 years was born of a culture that tests animals and interprets the results of those tests in attempts to protect public health. Even with the current awareness that there are many more chemicals to assess than resources and test data permit, those skeptical of QSAR are predominant in the regulatory setting. Bureaucracies founded on laboratory testing, whether a private or governmental agency, will only begrudgingly accept QSAR as a strategic tool for designing chemicals and managing chemical risks. Every major milestone in QSAR accomplishments has been met with stronger skepticism that QSAR cannot replace animal testing. The QSAR research community needs to embrace the arguments of the skeptics and design research to overcome the perceived inadequacies of current QSAR methods. This paper will discuss three common errors in QSAR research, which, if corrected, will place in silico methods fully complementary to the strategic use of in vitro and in vivo methods. QSAR is based on well-defined endpoints of intrinsic chemical activities and molecular descriptors, which can be mechanistically interpreted. Chemicals in a QSAR training set ought to have a common mechanism of interaction so that the context of structural requirements defining the domain can be articulated and tested. Finally, the estimation of complex endpoints ought to be controlled by a QSAR-based expert system if the estimation of missing values or hazard screening in heterogeneous inventories is to avoid fueling the skepticism of QSAR.