Analyzing gene expression data in terms of gene sets: methodological issues

Abstract

Motivation: Many statistical tests have been proposed in recent years for analyzing gene expression data in terms of gene sets, usually from Gene Ontology. These methods are based on widely different methodological assumptions. Some approaches test differential expression of each gene set against differential expression of the rest of the genes, whereas others test each gene set on its own. Also, some methods are based on a model in which the genes are the sampling units, whereas others treat the subjects as the sampling units. This article aims to clarify the assumptions behind different approaches and to indicate a preferential methodology of gene set testing. Results: We identify some crucial assumptions which are needed by the majority of methods. P-values derived from methods that use a model which takes the genes as the sampling unit are easily misinterpreted, as they are based on a statistical model that does not resemble the biological experiment actually performed. Furthermore, because these models are based on a crucial and unrealistic independence assumption between genes, the P-values derived from such methods can be wildly anti-conservative, as a simulation experiment shows. We also argue that methods that competitively test each gene set against the rest of the genes create an unnecessary rift between single gene testing and gene set testing. Contact: j.j.goeman@lumc.nl

Diffuse large B-cell lymphomas with CDKN2A deletion have a distinct gene expression signature and a poor prognosis under R-CHOP treatment: a GELA study

“…Hierarchic clusterings were performed by the Ward agglomeration algorithm. The gene expression profile observed in DLBCL 9p21del cases was analyzed using 2 statistical enrichment analyses: (1) global differences in biologic processes and signaling pathways between DLBCL 9p21del cases and others were studied with the Goeman Globaltest approach using biologic pathways provided by the Molecular Signatures Database (MSigDB) and molecular signatures from the lymphoma literature; 22,23 (2) functional analysis to identify the most relevant biologic mechanisms, pathways, and functional categories in the datasets of genes selected by statistical analysis were generated through the use of Ingenuity Pathways Analysis (IPA 6.5 software; Ingenuity Systems).…”

CD8‐alpha T‐cell infiltration in human papillomavirus‐related oropharyngeal carcinoma correlates with improved patient prognosis

“…8 Four methods were used to compare gene sets and sample groups: GSA 20 : R package GSA; globaltest 21 : R package globaltest; SAM-GS 22 : original R code; and the Tuckey algorithm described in Table 4 of Ref. 23: original R code). Each method yielded a p-value: the lower the p-value, the more the genes in this gene set are differentially expressed between the sample groups.…”

Data Integration and Knowledge Discovery in Life Sciences

“…It tests whether the rank of genes ordered according to P-values differs from a uniform distribution. Goeman and Bühlmann recently reviewed existing GSA methods, and strongly recommended the use of selfcontained methods [28]. We are currently developing and testing statistical methods for GSA analysis of cancer expression profiles.…”