You are here:
STRUCTURE-ACTIVITY RELATIONSHIP STUIDES AND THEIR ROLE IN PREDICTING AND INVESTIGATING CHEMICAL TOXICITY
Citation:
Richard, A M. STRUCTURE-ACTIVITY RELATIONSHIP STUIDES AND THEIR ROLE IN PREDICTING AND INVESTIGATING CHEMICAL TOXICITY. Presented at Symposium on Chemical Mechanisms of Toxicity: Basic Knowledge for Designing Safer Chemicals, Basel, Switzerland, May 4, 2000.
Description:
Structure-Activity Relationship Studies and their Role in Predicting and Investigating Chemical Toxicity
Structure-activity relationships (SAR) represent attempts to generalize chemical information relative to biological activity for the twin purposes of generating insight into the mechanism of activity and predicting the potential activities of new chemicals. In addition, assumption of a common mechanism of action or rate determining step, through restriction of chemical class or biological function, is a central guiding assumption that both constrains and enables development of meaningful SAR associations. A wide range of SAR investigations at different levels of resolution are considered. Global SAR models attempt to analyze structurally and mechanistically diverse data relative to a common toxicity endpoint (e.g. statistical programs such as CASE and TOPKAT). Since little or no bias or prior knowledge is assumed, these models are useful for data exploration and hypothesis generation. At the next level of resolution are the traditional QSAR or Hansch-type methods which attempt to derive quantitative models of relative potency in relation to structure, with the prior assumption that a common mechanism of action applies to the set of chemicals. The most detailed level of SAR study assumes some molecular-level resolution of the problem, relies on a clear prior hypothesis, and applies computational chemistry techniques to model key reactivity or interaction steps in a putative mechanism of action. Through a series of examples at each level of resolution, this presentation attempted to illustrate the wide range of toxicity problems that can be considered with SAR methods, the nature of the questions, and the types of information that can be gained from such studies. Issues pertaining to the use of global SAR models, such as the commercial toxicity prediction systems CASE and TOPKAT , stress the need to explore the structural basis for predictions and to move beyond statistics to build independent scientific rationale for models and individual predictions. Development of a modified CASE-FDA model for carcinogenicity prediction served to illustrate the strong dependence of the SAR model performance on the composition and size of the training set, the definition of activity, and the model assumptions. An example of developing SAR models for rat nasal toxicity served to illustrate how careful application of quantitative methods can yield useful information even with a small dataset for a chronic in vivo endpoint that is seemingly far from the ideal for SAR study. Models relative to haloacetic acids and halogenated alkenes served to illustrate how computational chemistry methods can be brought to bear in a very focused way to explore possible chemical and reactivity mechanisms underlying a biological activity. Finally, modelling of androgen and estrogen receptor binding were offered as examples of receptor-mediated events relying on 3D chemical information, that may or may not pertain to the ultimate toxicity endpoint of concern, and that mayor may not require explicit consideration and knowledge of the detailed receptor binding domain and ligand interaction. The talk concluded with a few comments relative to the challenges of applying SAR modeling of toxicity in the phm-maceutical industl-y , and to the evolving role of SAR in the new era of genomics, as a means for refining our understanding of chemical structure in relation to biology function.
This abstract does not necessarily reflect EPA policv.