ABSTRACT
In general, the model-building approaches are divided into two categories, namely, linear and nonlinear, which are based on the nature of relationship between the selected parameters and biological activities. Since the beginning, the linear methods in the quantitative structure-activity relationship (QSAR) analyses have been utilized extensively. This is mostly because of easy interpretation, simplicity, and reproducibility of the generated models. The simplest form of linear regression is called simple linear regression (SLR). It is based on a two-dimensional equation between dependent and independent variables. Multiple linear regression is an extension of SLR in which more independent variables are involved in a hyperdimensional space. The application of partial least square becomes essential when a large number of multicollinear descriptors are used for QSAR analyses. Linear discriminant analysis works based on classification of the biological activity values and then representing them in class label-based categorical values, but the molecular parameters are in the form of continuous values.
