Random-set methods identify distinct aspects of the enrichment signal in gene-set analysis

• DOI: 10.1214/07-aoas104
• Publication Date: 2007-08-22T21:56:33Z
• Authors: Michael A. Newton; Fernando A. Quintana; Johan A. den Boon; Srikumar Sengupta; Paul Ahlquist
• Keywords: Expression (computer science); False Discovery Rate; Multiple comparisons problem

A prespecified set of genes may be enriched, to varying degrees, for that have altered expression levels relative two or more states a cell. Knowing the enrichment gene sets defined by functional categories, such as ontology (GO) annotations, is valuable analyzing biological signals in microarray data. common approach measuring cross-classifying according membership category and on selected list significantly genes. small Fisher's exact test p-value, example, this 2×2 table indicative enrichment. Other analysis methods retain quantitative gene-level scores measure significance referring category-level statistic permutation distribution associated with original differential problem. We describe class random-set scoring distinct components signal. The includes based also tests average evidence across category. Averaging selection are compared empirically using Affymetrix data nasopharyngeal cancer tissue, theoretically location model expression. find each method has domain superiority state space problems, both benefits practice. Our addresses problems related multiple-category inference, namely, equally enriched categories not detected equal probability if they different sizes, there dependence among statistics owing shared Random-set calculations do require Monte Carlo implementation. They made available R package allez.