Extended Iterative Nonlinear Regression Normalization for cDNA Gene Expression Data

cDNA microarray expression data is widely used to help biomedical research. The data must be normalized because of various error functioned interferences existed. This paper has discussed the normalization for supervised multi-class (phenotype) data. All the classes are the type of multi-sample. Also, a reasonable hybrid cross-phenotype normalization (CPN) algorithm based on iterative nonlinear regression (INR) is proposed for this kind of array data set. As a part of this CPN algorithm, how to obtain a ldquobaselinerdquo from samples within a class by a statistical way and dynamic decision of reference/floating sample are discussed. Finally, experimental result is presented. The method in this paper has practical significance. Specifically, it can be used as a novel feature selection in gene pattern recognition.