A weighted hypergeometric statistic for the enrichment of gene sets

Microarray data can be analyzed by observing the activity of groups of genes; this approach can provide more relevant information than the analysis of individual gene expression. There are numerous statistical methods available for the enrichment of gene sets; however, these methods often fail to identify relevant gene sets due to the noisy nature of the data. We present weighted hypergeometric and weighted chi-squared methods that rank each gene's contribution to enrichment by the absolute value of the logarithm of its fold change. We demonstrate that these methods can produce more biologically relevant results than the standard hypergeometric test, despite being more conservative and enriching fewer pathways with significant p-values.