Gene Selection using the GMM-IG Framework based Integrative Analysis

The limitation of sample numbers is becoming a bottle neck in gene expression research. How to integrate the DNA microarray data involving in same or similar biological subjects is a promising way to increase sample numbers. However, the noise and disparities between different microarrays still make integration a challenge. Here we provide a straightforward and easily implemented framework to combine information from multiple microarrays. This framework applies a two-component Gaussian mixture modeling (GMM) to estimate the underlying expression levels of genes and discretize the original continuous values. Then significantly differentially expressed genes are selected and ranked by integrated microarray data and Information-Gain (IG) measure. The real data evaluation showed that this method performed better than other existing integrative methods.