Automatic preprocessing of electrophoretic images

Abstract Analysis of two-dimensional (2D) electrophoretic images is a multi-step approach, enabling application of a variety of methods at different stages of data processing. The choice of these, as well as input parameters, leads to software-induced variations. Effective preprocessing methods, which do not require optimization of input parameters, are potent in eliminating software-induced variations. As a general method for background elimination and image scaling, robust Orthogonal Regression (rOR) is proposed and compared with Orthogonal Regression. This comparison is based on the univariate and multivariate approaches of feature selection, exploring the idea developed for significance analysis of microarray data [V. Goss Tusher, R. Tibshirani, G. Chu, Significance analysis of microarrays applied to the ionizing radiation response, P. Natl. Acad. Sci. U. S. A., 98 (2001) 5116–5121] and adapted to the analysis of proteomic data. All calculations are performed at the pixel level.