Symmetry considerations in Markovian chemicals 'in silico' design (MARCH-INSIDE) I: central chirality codification, classification of ACE inhibitors and prediction of sigma-receptor antagonist activities.

The MARCH-INSIDE methodology has been generalized, by means of an exponential central symmetry factor, to codify chemical structure information for chiral drugs. In order to test the potential of this novel approach in drug design we have modeled the angiotensin-converting enzyme inhibitory activity of perindoprilate's sigma-stereoisomer combinatorial library. A linear discriminant analysis (LDA) model classifies correctly 83.33% of active compounds and 94.12% of non-active ones in a training set, results that represent a total of 91.3% accuracy in classification. On the other hand, the model classifies 83.33% of these compounds in the predicting series. Only three isomers (those with higher activity) were used in the predicting set and the model classified all three very well. Similar predictive behavior was observed in a leave-1-out cross validation experiment. Canonical regression analysis corroborated the statistical quality of the models (Rcanc=0.79, with a P-level<0.000) and was also used to compute biological activity canonical scores for each compound. Finally, prediction of the biological activities of chiral 3-(3-hydroxyphenyl)piperidines, which are sigma-receptor antagonists, by linear regression analysis was carried out. The model was statistically significant (R=0.963, S=0.29, P<0.00) and can be considered as a preliminary comparative study between MARCH-INSIDE and Chiral Topologic descriptors. Application of the Student test permits the detection of non-symmetric properties within the data set and justified the requirement of non-symmetric (for pairs of enantiomers) molecular descriptors. The MARCH-INSIDE model showed very good stability to data variation in the leave-1-out cross validation experiment (Scv=0.32).