Dataset Information

Should we abandon the t-test in the analysis of gene expression microarray data: a comparison of variance modeling strategies.

ABSTRACT: High-throughput post-genomic studies are now routinely and promisingly investigated in biological and biomedical research. The main statistical approach to select genes differentially expressed between two groups is to apply a t-test, which is subject of criticism in the literature. Numerous alternatives have been developed based on different and innovative variance modeling strategies. However, a critical issue is that selecting a different test usually leads to a different gene list. In this context and given the current tendency to apply the t-test, identifying the most efficient approach in practice remains crucial. To provide elements to answer, we conduct a comparison of eight tests representative of variance modeling strategies in gene expression data: Welch's t-test, ANOVA [1], Wilcoxon's test, SAM [2], RVM [3], limma [4], VarMixt [5] and SMVar [6]. Our comparison process relies on four steps (gene list analysis, simulations, spike-in data and re-sampling) to formulate comprehensive and robust conclusions about test performance, in terms of statistical power, false-positive rate, execution time and ease of use. Our results raise concerns about the ability of some methods to control the expected number of false positives at a desirable level. Besides, two tests (limma and VarMixt) show significant improvement compared to the t-test, in particular to deal with small sample sizes. In addition limma presents several practical advantages, so we advocate its application to analyze gene expression data.

SUBMITTER: Jeanmougin M

PROVIDER: S-EPMC2933223 | biostudies-literature | 2010 Sep

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Should we abandon the t-test in the analysis of gene expression microarray data: a comparison of variance modeling strategies.

Jeanmougin Marine M de Reynies Aurelien A Marisa Laetitia L Paccard Caroline C Nuel Gregory G Guedj Mickael M

PloS one 20100903 9

High-throughput post-genomic studies are now routinely and promisingly investigated in biological and biomedical research. The main statistical approach to select genes differentially expressed between two groups is to apply a t-test, which is subject of criticism in the literature. Numerous alternatives have been developed based on different and innovative variance modeling strategies. However, a critical issue is that selecting a different test usually leads to a different gene list. In this c ...[more]

PMID: 20838429

Similar Datasets

Project description:BackgroundMany microarray experiments search for genes with differential expression between a common "reference" group and multiple "test" groups. In such cases currently employed statistical approaches based on t-tests or close derivatives have limited efficacy, mainly because estimation of the standard error is done on only two groups at a time. Alternative approaches based on ANOVA correctly capture within-group variance from all the groups, but then do not confront single test groups with the reference. Ideally, a t-test better suited for this type of data would compare each test group with the reference, but use within-group variance calculated from all the groups.ResultsWe implemented an R-Bioconductor package named Mulcom, with a statistical test derived from the Dunnett's t-test, designed to compare multiple test groups individually against a common reference. Interestingly, the Dunnett's test uses for the denominator of each comparison a within-group standard error aggregated from all the experimental groups. In addition to the basic Dunnett's t value, the package includes an optional minimal fold-change threshold, m. Due to the automated, permutation-based estimation of False Discovery Rate (FDR), the package also permits fast optimization of the test, to obtain the maximum number of significant genes at a given FDR value. When applied to a time-course experiment profiled in parallel on two microarray platforms, and compared with two commonly used tests, Mulcom displayed better concordance of significant genes in the two array platforms (39% vs. 26% or 15%), and higher enrichment in functional annotation to categories related to the biology of the experiment (p value < 0.001 in 4 categories vs. 3).ConclusionsThe Mulcom package provides a powerful tool for the identification of differentially expressed genes when several experimental conditions are compared against a common reference. The results of the practical example presented here show that lists of differentially expressed genes generated by Mulcom are particularly consistent across microarray platforms and enriched in genes belonging to functionally significant groups.

Project description:Schizophrenia (SCZ) is a psychiatric disorder of unknown etiology. There is evidence suggesting that aberrations in neurodevelopment are a significant attribute of schizophrenia pathogenesis and progression. To identify biologically relevant molecular abnormalities affecting neurodevelopment in SCZ we used cultured neural progenitor cells derived from olfactory neuroepithelium (CNON cells). Here, we tested the hypothesis that variance in gene expression differs between individuals from SCZ and control groups. In CNON cells, variance in gene expression was significantly higher in SCZ samples in comparison with control samples. Variance in gene expression was enriched in five molecular pathways: serine biosynthesis, PI3K-Akt, MAPK, neurotrophin and focal adhesion. More than 14% of variance in disease status was explained within the logistic regression model (C-value = 0.70) by predictors accounting for gene expression in 69 genes from these five pathways. Structural equation modeling (SEM) was applied to explore how the structure of these five pathways was altered between SCZ patients and controls. Four out of five pathways showed differences in the estimated relationships among genes: between KRAS and NF1, and KRAS and SOS1 in the MAPK pathway; between PSPH and SHMT2 in serine biosynthesis; between AKT3 and TSC2 in the PI3K-Akt signaling pathway; and between CRK and RAPGEF1 in the focal adhesion pathway. Our analysis provides evidence that variance in gene expression is an important characteristic of SCZ, and SEM is a promising method for uncovering altered relationships between specific genes thus suggesting affected gene regulation associated with the disease. We identified altered gene-gene interactions in pathways enriched for genes with increased variance in expression in SCZ. These pathways and loci were previously implicated in SCZ, providing further support for the hypothesis that gene expression variance plays important role in the etiology of SCZ.

Dataset Information

Should we abandon the t-test in the analysis of gene expression microarray data: a comparison of variance modeling strategies.

Publications

Should we abandon the t-test in the analysis of gene expression microarray data: a comparison of variance modeling strategies.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets