Traditional analysis of gene expression data focused on identifying differentially expressed and co-expressed genes, which didn′t take known interactions into consideration. In recent years, many methods have been developed to identify active subnetwork by integrating protein interaction networks with gene expression profiles. Current approaches failed to take full account of both difference and correlation in gene expression that may lead to false positive results. A new algorithm is proposed for identifying active subnetwork by considering both difference and correlation of gene expression profile. The algorithm is employed in the process of gene expression profiles of human immunodeficiency virus infection. The results showed that the algorithm can identify the active subnetwork that has extremely high biologically functional connectivity with human immunodeficiency virus, and effectively avoid the bias introduced by considering differences of gene expression profiles only, i.e., genes less differentially expressed are also included due to high correlations in gene expression.