ABSTRACT
Spectral-SAR (S-SAR) represents a new approach to study quantitative structure-activity relationships, in an exclusively algebraic manner, replacing the old multi-regressional model. This method highlights structural descriptors as vectors in a generic data space using the Gram-Schmidt algorithm. Applicability of methods such as S-SAR (analytical equations) lies in the similarity of obtained results when reference is made to standard multivariate statistical correlation results. Although the main goal of QSAR studies is to find structural parameters that correlate best way possible with the activity/property of targeted interactions, there is a variety of mathematical and computational methods for achieving this. In other words, QSAR seeks to identify the most appropriate way of quantifying the causes so that they can be reflected in measurements with maximum accuracy and minimum errors. S-SAR method provides a real alternative to the standard QSAR at every level of modeling, independently of the number of descriptors, compounds or of the order of orthogonalization of considered parameters.
