A variance component method for integrated pathway analysis of gene expression data

BACKGROUND: The application of pathway and gene-set based analyses to high-throughput data is increasingly common and represents an effort to understand underlying biology where single-gene or single-marker analyses have failed. Many such analyses rely on the a priori identification of genes associated with the trait of interest. In contrast, this variance-component–based approach creates a similarity matrix of individuals based on the expression of genes in each pathway. METHODS: We compared 16 methods of calculating similarity for positive control matrices based on probes for the genes used to model the simulated Genetic Analysis Workshop phenotypes. RESULTS: A simple correlation matrix outperforms the other methods by identifying pathways associated with the simulated phenotypes at nearly twice the rate expected based on the associations of the component transcripts and an approximate false-positive rate of 0.05. CONCLUSIONS: This method has a number of additional advantages compared to single-transcript and pathway overrepresentation analyses, including the ability to estimate the proportion of variation explained by each pathway and the logistical advantage of only calculating the distance matrices once for each messenger RNA data set regardless of the number of phenotypes. Additionally, it offers a significant reduction in the multiple testing burden over individual consideration of each probe.