Systematic QSAR Procedures with Quantum Chemical Descriptors

Abstract

AbstractSystematic QSAR procedures with quantum chemical descriptors were described by showing practical calculations on the inhibition of thiopurine methyltransferase by benzoic acid derivatives. In the QSAR analysis, a large number of descriptors relating to reactivity and interaction efficiency are handled, and the collinearities inherently involved among them should be removed to get the stable regression equations. Collinearity was detected by the A value which is the sum of the reciprocals of the eigenvalues of correlation matrix of descriptors divided by the number of descriptors in the equation. In the method of evaluation of all possible equations, the candidate equations selected on the basis of the correlation coefficient with activity and/or predicted sum of square should be examined one by one by the A testing. In the stepwise methods, noncollinear descriptor set is required. The collinearity between two descriptors was removed by a simple method, deleting one of the two descriptors whose correlation is large. It was recommended that the A testing routine was introduced into each step to find out multicollinearity.