Application of orthogonal pixel correction for MIA-QSAR study of anilinquinolines

The aim of this study was to investigate, for the first time, the possibility of using orthogonal signal (pixel) correction (OPC) and multivariate calibration methods for QSAR study of anilinquinolines [1] by image analysis. The results obtained, with and without using OPC algorithm as a preprocessing treatment of original data, were compared. The aim of this work is to propose OPC and PLS (OPC-PLS) method to modeling in QSAR study. To our knowledge this is the first report on the QSAR study. In the MIA-QSAR [2, 3] method, the descriptors are the pixels of images that can be two or three dimensional. These pixels are correlated with dependent variables for making QSAR models [4]. Each image of dimension 120×150 pixels was unfolded to a 1×18000 row and then the 62 images were grouped to form a 62×18000 matrix. Columns with zero variance were removed to minimize memory, reducing the size of matrix to 62×1405. The resulted model showed high prediction ability with RMSEP of 0.0106 for OSC-PLS.