Normalization of cDNA microarray data by using neural networks

In microarray experiments, there are a variety of systematic errors to affect the measured gene expression levels. Although a number of algorithms were proposed for the normalization of different types of cDNA microarray data, they have encountered many difficulties due to the complex nonlinear sources of systematic error. In this case, a nonlinear normalization method is of great potential to deal with this difficult problem. The paper first proposes a nonlinear method for the normalization, i.e. neural network normalization (N3) approach, of cDNA microarray experiments in the community of bioinformatics. By utilizing the instinctive nonlinear processing ability of neural networks, N3 is able to balance the complex nonlinear dependence between two different dyed channels in cDNA microarray experiments. In such a way, we can obtain much better normalization performance of cDNA microarray data than current existing approaches. Several experiments are conducted to illustrate the validation of our proposed methods in detail