Parameter-free linear relationship (PFLR) and its application to 3D QSAR

摘要

The linear relationship is still the most important tool for establishing connection between correlating features, properties. The name “parameter-free linear relationship” (PFLR) stands for a new formalism, a generalized interpolation scheme, which can be readily used for predicting biological activities or other properties in 3D QSAR manner. Our studies demonstrate the good predictive power of PFLR even when used with a simple set of 3D molecular descriptors without constructing a grid representation of the features. PFLR allows completing most of the computations solely in the space of descriptors, without experimental training data, which, however, bears no importance in the case of 3D QSAR but might be advantageous in other areas where multidimensional linear regression or partial least squares based methods are applicable.