Dedicated principal properties for peptide QSARs: Theory and applications

Dedicated principal properties (DPPs) are presented. The DPP approach is an iterative process based on PLS wherein principal properties are calibrated on QSARs with improved predictive capabilities as a consequence. The DPP approach is illustrated by two examples. In the first example the z‐scales developed for structural description in peptide QSARs are optimized according to the DPP approach. In the second example DDPs based on simulated pentapeptide‐like sequences are calculated and evaluated. A comparison is made of the predictive power of peptide QSAR models based on structural description of the peptides with (a) 29 physico‐chemical variables with (b) three principal properties (z‐scales) and (c) three sets of DPPs. The results show that models based on all 29 variables give better predictions than models based on principal properties. Modelling based on DPPs improves the predictive power to the same level as models based on all 29 variables.