Inference of Gene Coexpression Networks by Integrative Analysis across Microarray Experiments

We improve the reliability of detecting coexpressed gene pairs from microarray data by introducing a novel probe-level quality-weighted similarity measure for combining data across different Affymetrix experiments. In construction of gene coexpression networks, the proposed procedure is less sensitive to noise than the corresponding single-experiment approaches or the conventional integrative approaches, even when a relatively small number of samples and conditions is available. The present results indicate how the accumulated microarray data can be effectively exploited to increase the quality of the inferred networks. In particular, we demonstrate its biological relevance in identifying coexpressions in mouse T helper cell differentiation.