Gene set enrichment analysis: performance evaluation and usage guidelines

Abstract

A central goal of biology is understanding and describing the molecular basis of plasticity: the sets of genes that are combinatorially selected by exogenous and endogenous environmental changes, and the relations among the genes. The most viable current approach to this problem consists of determining whether sets of genes are connected by some common theme, e.g. genes from the same pathway are overrepresented among those whose differential expression in response to a perturbation is most pronounced. There are many approaches to this problem, and the results they produce show a fair amount of dispersion, but they all fall within a common framework consisting of a few basic components. We critically review these components, suggest best practices for carrying out each step, and propose a voting method for meeting the challenge of assessing different methods on a large number of experimental data sets in the absence of a gold standard.

Principle Angle Enrichment Analysis (PAEA): Dimensionally reduced multivariate gene set enrichment analysis tool

“…We used the popular method of Gene Set Enrichment Analysis [8] along with a collection of methods listed by DeLisi and coworkers [7]. These methods are: the χ 2 test, mean p value test, median p value test, Wilcoxon rank sum test, and the weighed Kolmogorov-Smirnov (WKS) test.…”

“…Then, using this notation the, χ 2 gene set statistic is given by: $$\sum _{i\in \lambda }{false(p_i-<p>false)}^2$$ where < p > is the mean of { p i ∀ i ∈ λ}. The mean test gene set statistic is given by, $$\frac{1}{false|Lfalse|}\sum _{i\in \lambda }{p}_i$$ The median test gene set statistic is given by, $$true\{\begin{array}{cc}left{y}_{n/2}(L)& left\text{if }n\text{ is odd}\\ left\frac{1}{2}{y}_{n/2}(L)+y_{1+|L|/2}(L)& left\text{if}|L|\text{is even}\end{array}$$ The Wilcoxon rank sum test statistic is given by, $$\sum _{i\in \lambda }{\text{rank}}_{L+L\prime }false(p_{i}false)$$ And the weighted Kolmogorov Smirnov (WKS) test is given by the maximum deviation, taken over i, between, $$\mathrm{normalc}\mathrm{normald}{\mathrm{f}}_{p}false(ifalse)=\sum _{j\le i,j\in \lambda }\frac{y_{j}false(L+L\prime false)}{{\sum }_{k\in L}false(L+L\prime false)}$$ and $$\mathrm{normalc}\mathrm{normald}{\mathrm{f}}_{L\prime }false(ifalse)=\sum _{j\le i}\frac{1}{false|L\prime false|}$$ These are all as described by Hung et al [7]. We also compare the performance of these gene set level statistics when using the characteristic direction instead of the t test p values for the gene-level statistics.…”

“…However, synthetic data is only a pale imitation of real expression data, lacking the rich, yet poorly understood, structure present in real genome-wide expression data. DeLisi and coworkers [7] highlighted this challenge and devised a method of comparing methods of analysis whereby the degree to which a method is in consensus with other methods is taken as a measure of its performance quality. However, this measure of performance quality depends solely on the relationship of the methods to each other.…”

Integrated network analysis and effective tools in plant systems biology

“…Interaction-based approaches such as topological analysis (e.g., shortest path search Yu et al, 2014, centrality analysis Carrera et al, 2009, and network module detection Altaf-Ul-Amin et al, 2006), correlation network analysis (Provart, 2012), or enrichment analysis (Hung et al, 2012) have been used to construct and analyze biological networks from omics data. For example, GeneMANIA (Montojo et al, 2010; Zuberi et al, 2013) is a web-based interaction network for the visualization of physical, genetic, and functional interactions.…”

DBGSA: a novel method of distance-based gene set analysis

“…There are two primary types of gene function enrichment analysis methods, individual gene analysis (IGA) and gene set analysis (GSA). [1][2][3] IGA identifies differentially expressed genes through a variety of methods and tests the difference of the proportion of differentially expressed genes between all genes and a given gene set. 4,5 GSA directly calculates gene subset scores using various statistical methods and calculates the significance level.…”