Answers are grounded in this paper and cite their sources.
The impact of orthogonal signal correction (OSC) on the prediction power of CoMFA models was studied using a data set of 47 nitrobenzenes with toxicities (log 1/IC50) towards the aquatic ciliates Tetrahymena pyriformis. Comparative analyses of different data pre-treatments shows that block unscaled weighting (BUW) results in significantly better PLS models than no scaling, centering or autoscaling for OSC. One OSC component is optimal for the signal correction and reduces the X variance by about 40%. While OSC yields improved calibration and cross-validation statistics, standard CoMFA is superior with respect to the external prediction power as evaluated by models built from complementary subsets. Moreover, external prediction reveals some cases of severe OSC overfitting, which needs attention in future investigations.
No takes yet. Share an insight, caveat, or question.
Boháč et al. (2002) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: