QSAR modeling with the electrotopological state: TIBO derivatives.

Quantitative structure-activity relationships (QSAR), based on the atom level E-state indices and calculated molecular properties (log P, MR), have been developed for the affinity of a large set of TIBO derivatives against HIV-1 reverse transcriptase (HIV-1 RT) utilizing multiple linear regression techniques. A model with five descriptors, including four atom level E-state indices (carbon atoms 2, 4, 8, and 9) and calculated log P, showed good statistics both in the regression (r2 = 0.85 and s = 0.52) and leave-one-out cross-validation (q2 = 0.80 and s(PRESS) = 0.56) for the training set of 41 compounds. The statistics for the prediction of anti-HIV activity in the test set of 24 TIBO derivatives were r2 = 0.80 and s = 0.64, respectively. The model descriptors indicate the importance of lipophilic and electronic contributions toward HIV-1 RT inhibition of TIBO derivatives used in this study.