Abstract

DESCRIPTORS COMBINATION TO FOLLOW UP RESPONSE VARIABILITY OF SOME KNOWN ANTICANCER AMINOPYRIMIDOISOQUINOLINEQUINONES; A QSAR ANALYSIS

Mukhtaar Q. S. Sultan*, M. O. El-Fakii and Inas O. K. Mohammed

ABSTRACT

In search of a predictive QSAR model, various molecular descriptors were regressed against inhibitory concentration (IC50) of human gastric adenocarcinoma (AGS) cell line of 29 known aminopyrimidoisoquinolinequinones (APIQ). After data curation, 6 data points were removed from the data set and the remaining 23 data points fail to give valid simple linear correlations with separate descriptors. Depending on the overall correlation matrix of all the descriptors, we singled out the reciprocal of the difference of the heats of formation of the hydroquinone-Quinone forms (1/H) as the descriptor which gives the highest correlation with the response (R2=0.79) indicating that a thermodynamic connection is involved in the mechanism of action of APIQ`s. The validity of the resulted model was found to be weak owing to weakness of the cross-validation metric (Q2 = 0.023). Thus we run several regression analyses choosing from the correlation matrix the descriptors which give the highest correlation with the response and the lowest collinearity with each other. Various combination thereof were explored until a reasonable correlation was arrived at using (1/H), octanol/water partition coefficient (log P), molecular volume (MV) and electrophilicity index () (R2=0.9). This particular model shows high predictability in term of both internal and external validation metrics (Q2 = 0.6, R2pred=0.55 respectively). It is interesting to point that removal of any descriptor from the combination leads to drastic decrease in Q2 (from 0.6 to 0.1).

Keywords: Aminopyrimidoisoquinolinequinones APIQ`s, QSAR, Descriptors Combination.