SAR modeling: effect of experimental ambiguity.

The applicability of SAR (structure-activity relationship) techniques to data obtained using high throughput screening (HTS) and toxicogenomic techniques is explored. The reason for this study derives from the fact that for economical and time considerations HTS bioassays may consist of single determinations, i.e. lack of duplication. This introduces an element of uncertainty. Using two different data bases of fairly complex biological phenomena (allergic contact dermatitis in humans and the induction of mutations in Salmonella), it is demonstrated that the resulting SAR models can tolerate up to 20% ambiguity in the experimental data.