Enrichment map: a network-based method for gene-set enrichment visualization and interpretation.
Merico D, Isserlin R, Stueker O, Emili A, Bader GD. · 2010
Read Enrichment map: a network-based method for gene-set enrichment visualization and interpretation. by Merico D, Isserlin R, Stueker O, Emili A, Bader GD. (2010) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/europmc-pmc2981572.txt.
Introduction
Merico, Isserlin, Stueker, Emili, and Bader introduced Enrichment Map in 2010, a visual method for making sense of gene-set enrichment results. Such analyses flag whole pathways over-represented in a gene list, yet overlapping sets pile up into a redundant thicket.
Their tool draws each gene set as a node and links sets that share many genes, letting layout software gather related sets into clusters. Users can spot the main biological themes at a glance instead of reading long ranked tables. Released as a free plug-in for a popular network viewer, it suits any project that produces gene lists.
Catalog record
| CREATOR | Merico D, Isserlin R, Stueker O, Emili A, Bader GD. |
|---|---|
| YEAR | 2010 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 57 KB |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Enrichment Map: A Network-Based Method for Gene-Set Enrichment Visualization and Interpretation By Daniele Merico, Ruth Isserlin, Oliver Stueker, Andrew Emili, Gary D. Bader, Timothy Ravasi Published in: PLoS ONE DOI: 10.1371/journal.pone.0013984 ABSTRACT Background Gene-set enrichment analysis is a useful technique to help functionally characterize large gene lists, such as the results of gene expression experiments. This technique finds functionally coherent gene-sets, such as pathways, that are statistically over-represented in a given gene list. Ideally, the number of resulting sets is smaller than the number of genes in the list, thus simplifying interpretation. However, the increasing number and redundancy of gene-sets used by many current enrichment analysis software works against this ideal. Principal Findings To overcome gene-set redundancy and help in the interpretation of large gene lists, we developed “Enrichment Map”, a network-based visualization method for gene-set enrichment results. Gene-sets are organized in a network, where each set is a node and edges represent gene overlap between sets. Automated network layout groups related gene-sets into network clusters, enabling the user to quickly identify the major enriched functional themes and more easily interpret the enrichment results. Conclusions Enrichment Map is a significant advance in the interpretation of enrichment analysis. Any research project that generates a list of genes can take advantage of this visualization framework. Enrichment Map is implemented as a freely available and user friendly plug-in for the Cytoscape network visualization software (http://baderlab.org/Software/EnrichmentMap/). INTRODUCTION High-throughput genomic experiments often lead to the identification of large gene lists [1], [2], [3]. Gene lists are typically defined using statistical methods appropriate to the experimental design. For instance, a frequently applied method is to score genes by their differential expression between two biological states (such as healthy vs. diseased). Especially in more mature fields, like gene expression microarrays, the statistical models used for gene scoring are well established [1], [4]. For instance, gene expression values are ranked to identify the top-most list of expressed genes, based on an arbitrary expression threshold, or a set of gene expression experiments are clustered, each cluster defining a potentially large gene list. However, these methods for finding interesting genes often do not help the interpretation of the resulting gene lists and the formulation of consistent biological hypotheses from these results still poses a major challenge for experimentalists. Searching for sets of predefined functionally related genes (e.g. pathways) that are enriched in a gene list is a popular method designed to solve this problem. However, as gene-set collections get larger and more complex, users may experience longer lists of results and increased redundancy between sets. We have developed a visualization method for gene set enrichment results, called Enrichment Map, which helps quickly find general functional themes in genomics data. In the next sections, we introduce enrichment analysis, the gene set redundancy problem and then explain how Enrichment Map works using typical analysis scenarios. Enrichment analysis Early approaches to gene list interpretation relied on choosing a handful of high scoring genes, and then building rather subjective, anecdotal interpretations. Enrichment analysis is an automated and statistically rigorous technique to analyze and interpret large gene lists using a priori-knowledge [5]. Enrichment analysis assesses the over- (or under-) representation of a known set of genes (e.g. a biological pathway) within the input gene list [6], [7], [8]. If a statistically significant number of genes from the known set are present in the gene list, it may indicate that the biological pathway plays a role in the biological condition under study. This analysis is repeated for all available known gene-sets, which could number in the thousands. Over 60 enrichment analysis methods and tools have been developed in the last few years [5], [9], [10]. They mainly differ in (a) their database of known gene-sets and (b) the statistical method used to assess enrichment. In the following paragraphs, we briefly review existing approaches for enrichment analysis considering these two facets. Most enrichment tools derive gene-sets from Gene Ontology (GO) annotations [11], because they are readily accessible for many organisms and cover many genes, yet many other sources of gene-sets exist and are used by some tools in addition to GO [7], [12], [13]. Gene-sets can be defined based on participation in a metabolic or signaling pathway (e.g. KEGG [14], Reactome [15]), targeting by gene expression regulators (e.g. microRNA, transcription factors), protein features such as domains, chromosomal location and association to specific diseases, stimuli, or genetic perturbations. Gene-sets from multiple sources are collected in resources such as MSigDB [13] or WhichGenes [16]. Not all organisms are well covered by gene-sets and many tools only support specific organisms. Statistical methods to determine enrichment are usually either threshold-dependent or whole-distribution [9]. Threshold-dependent techniques require the user to input a discrete list of top-ranking genes, which may require setting a threshold on the gene scoring statistic. The one-tail Fisher's Exact Test [17], based on the hypergeometric distribution, was the first method proposed to address this problem [18], and continues to be one of the most used testing methods of this type [19]. These methods are useful for naturally discrete lists, but have major drawbacks when utilized with continuous gene scores. Specifically, results may not be stable to choice of threshold [20], and there is loss of information caused by treating gene scores in a binary way (they either pass the threshold or not). On the other hand, whole-distribution methods are threshold-free, as they test gene-sets by comparing their score distribution versus the background distribution. For this reason they are often preferred over threshold-dependent methods for gene lists associated with a continuous score. GSEA (Gene-Set Enrichment Analysis) [13], which utilizes the gene rank derived from differential expression or other statistics, is one of the most popular techniques in this group, though other whole-distribution testing models have been proposed [5], [9]. The gene-set redundancy problem The growing number of available gene-sets, due to the increased availability of functional annotations, makes enrichment analysis a powerful tool to help researchers gain interesting insights from their high-throughput data. However, this comes at a cost: as gene-set collections get larger and more complex, there may be longer lists of results and increased redundancy between sets. Redundancy is particularly problematic with gene-sets derived from hierarchical functional annotation systems, like GO, as children terms are partially redundant with their parents by definition. Gene-set redundancy constitutes a major barrier for the interpretability of enrichment results, limiting the full exploitation of its analytic power. This problem can be addressed by modifying either the statistical test or the gene-sets to minimize the effect of redundancy and produce more concise enrichment results. Existing methods usually take advantage of the hierarchical structure of Gene Ontology to reduce redundancy, a solution that is only effective for GO or other hierarchically organized gene sets and not applicable to many others, such as pathways, experimental signatures and regulator targets. POSOC [21] exploits the GO hierarchy to merge single sets into clusters, which can then be tested for enrichment. Ontologizer [22] defines a modified Fisher's Exact Test for hierarchical vocabularies, termed the parent-child approach. Enrichment of a given gene-set is calculated with respect to the parent gene-set, instead of the list of genes in the experiment (the experimental universe set), in order to downplay uninformative child enrichment