InteractiVenn: a web-based tool for the analysis of sets through Venn diagrams
© Heberle et al.; licensee BioMed Central. 2015
Received: 28 January 2015
Accepted: 6 May 2015
Published: 22 May 2015
Set comparisons permeate a large number of data analysis workflows, in particular workflows in biological sciences. Venn diagrams are frequently employed for such analysis but current tools are limited.
We have developed InteractiVenn, a more flexible tool for interacting with Venn diagrams including up to six sets. It offers a clean interface for Venn diagram construction and enables analysis of set unions while preserving the shape of the diagram. Set unions are useful to reveal differences and similarities among sets and may be guided in our tool by a tree or by a list of set unions. The tool also allows obtaining subsets’ elements, saving and loading sets for further analyses, and exporting the diagram in vector and image formats. InteractiVenn has been used to analyze two biological datasets, but it may serve set analysis in a broad range of domains.
InteractiVenn allows set unions in Venn diagrams to be explored thoroughly, by consequence extending the ability to analyze combinations of sets with additional observations, yielded by novel interactions between joined sets. InteractiVenn is freely available online at: www.interactivenn.net.
KeywordsVenn diagram Edwards-Venn diagram Interaction Set-unions
In biological sciences it is often necessary to compare sets of data such as genes, proteins, organisms as well as other entities. The same set comparison activities can be useful to many other research fields. Sets and their unions and intersections may be conveniently displayed as Venn diagrams, a widely adopted and familiar layout . A recent example can be seen in the article describing the banana genome , where a Venn diagram illustrates the relationship among six plant genomes. Venn diagrams may also be used as a means to explore and reason about data, but most often such diagrams provide only static views of up to four sets of data.
Although Venn diagrams may be built for any number of sets, the layout becomes increasingly challenging beyond four sets. For instance, we can use circles to create a symmetric three-set diagram but we cannot use them with four or more sets and still show all possible intersections and exclusive regions. The same holds for ellipses: we cannot create a diagram with more than five sets with them. Many other shapes may be used, such as squares, triangles and spherical surface segments, the latter being a widely adopted layout proposed by Edwards . A Venn diagram could even be formed by elements with abstract shapes, but not without difficulties in showing all necessary regions and in generalizing for any number of sets.
With a larger number of sets, symmetric Venn diagrams are easier to interpret because the reader needs less time to locate regions of interest and their boundaries, as well as set intersections. Considering restrictions in aesthetics and burden on human visual ability, Venn diagrams of seven sets just seem to be too much.
Many websites and graphics programs allow users to manually draw labeled Venn diagrams. Other programs extend the basic drawing, automating the construction of diagrams from lists of elements in each set. Examples include Pangloss Venn diagram generator (http://www.pangloss.com/seidel/Protocols/venn4.cgi) and Venny (http://bioinfogp.cnb.csic.es/tools/venny/), both restricted to static diagrams of up to four sets. Other examples are DrawVenn (http://bioinformatics.psb.ugent.be/webtools/Venn/) and BioVenn  that construct area-proportional diagrams, limited by the fact that it is not possible to create symmetric Venn diagrams with more than four sets that respect proportionality. Software systems that build area-proportional diagrams for larger number of sets use heuristics and typically produce Euler diagrams without support for all possible intersections among sets.
However, these choices are limited either in the number of sets they handle or in the ability to display partial set operations (such as partial unions inside the diagram). Displaying partial unions, that is, locating regions that combine unions of sets and their intersections, and highlighting such regions interactively, can provide additional insight to the analyst. In this article we describe InteractiVenn, a web-based tool to build and analyze Venn diagrams. InteractiVenn provides the ability to interactively visualize the effect of successive unions of up to six sets, enhancing the user capacity of analyzing data. Additionally, set unions are allowed in any order, including along a union list or a binary tree, enabling different perspectives on the data, for instance in the analysis of sets of genes across species related by a phylogeny.
Results and discussion
When the user moves the cursor over the diagram, the set under the cursor will be highlighted to help identify its intersections. Transparency is employed to help locate intersections and ease the identification of each region in the diagram. Pressing a mouse button on a non-empty region triggers the display of a list of the elements in the intersection of sets represented by that region.
InteractiVenn enables unions of sets preserving the diagram shape. Such unions of sets are often useful to unveil differences and similarities among them. The unions may be guided either by a binary tree or by a list. The selection between tree or list is made via the main toolbar of InteractiVenn.
A more general union specification may be given by a list of sets to unite at each step. The navigation between diagrams obtained through the tree in Fig. 2 may be also obtained specifying the list ;C D;C D B,C D B A. The list allows defining unions that are not hierarchically related, for instance ;C D;C B;A B,C D will produce a diagram where no sets are united, then one diagram with the sets A, B and C∪D, another diagram with A, C∪B and D, and finally a diagram with the sets A∪B and C∪D.
InteractiVenn allows adjusting font size as well as setting and opacity, that may be used by the user to tune the diagram layout. It also allows exporting the diagram in PNG and SVG formats, and the elements in every diagram region may be exported as a text file. InteractiVenn allows saving the sets to a local file that can be uploaded in a future moment for further analyses.
Study case 1: comparing feature selection methods for candidate biomarkers discovery
To show the usefulness of our tool, we have analyzed a published prostate cancer proteomic dataset , searching for candidate biomarkers through feature selection analyses. In a previous work by Kawahara et al. , a discovery-to-target pipeline was proposed to analyze proteomics data, comprising a mass spectrometry (MS)-based discovery, three feature selection methods, clustering, Venn diagram, bioinformatics analyses and targeted approaches. The feature selection methods used in the pipeline were the univariate Beta-binomial , the semi-multivariate Nearest Shrunken Centroids (NSC)  and the multivariate Support Vector Machine-Recursive Features Elimination (SVM-RFE) . The proof-of-concept was performed in a well-controlled proteomic data from the secretomes of three human cell lines, and was also validated on the published prostate cancer proteomic dataset .
Here, in order to generate lists of proteins sorted by relevance in discriminating the two classes in the dataset (organ-confined (OC) and extracapsullar (EC) prostate cancer cells), five methods were used, including the three used before in the discovery-to-target pipeline , the classical t test and the MWW test, all implemented in R. These and many other methods are being developed and applied for this type of study and they need to be compared for a deeper understanding on the distribution of candidate biomarkers resulting from the different methods.
As described by Tibshirani et al. , we may consider the top-n proteins to look for potential biomarkers. Although there are statistical validation procedures to define the value of n and to calculate false-positive and false-negative rates for each method , we adopt the vision that it is also important to compare all resulting lists of proteins, since those that appear in most methods may be more reliable or may lead to less false-positives, and thus can be further used by biologists in future experiments. Moreover, in a second analysis step, we may compare both the intersecting and exclusive proteins of each method to determine if one is good for potential biomarkers identification.
Furthermore, seven proteins identified as candidate biomarkers in the prostate cancer cells in the work by Kim et al.  were also verified in the same work by experimental biochemical methods and were searched in the Venn diagram sets built using the InteractiVenn tool: KLK3 (PSA), ACPP (PAP), SFN, MME, PARK7, TIMP1 e TGM4. Notably, from these proteins, KLK3 was the only one not validated as a candidate biomarker and, using InteractiVenn, we could observe that it was retrieved as an exclusive protein only by the MWW test (Fig. 3C). Out of the other six validated candidates, four (ACPP, SFN, MME e TGM4) were found in the intersection among the three methods used in the discovery-to-target pipeline , one (PARK7) was found in the intersection between Beta-binomial and NSC, and another one (TIMP1), in the intersection between NSC and SVM-RFE. Interestingly, none was found exclusively by the t test, suggesting that the three methods used in the pipeline described by Kawahara et al.  could retrieve the best potential candidate biomarkers in their intersections.
Study case 2: distribution of gene families among six plant genomes
A Venn diagram comparing six plant proteomes is shown in the article describing the banana (Musa acuminata) genome and the evolution of monocotyledonous plants . The diagram shows the distribution of shared gene families (sequence clusters) among the proteomes of five monocotyledons (Musa acuminata, Phoenix dactylifera, Oryza sativa, Sorghum bicolor and Brachypodium distachyon) and one eudicotyledon (Arabidopsis thaliana), and it was used to compare how many sequence clusters exist in common among these species.
As an example, Fig. 4A shows a simplified tree that represents the hierarchy that appears in the M. acuminata article, with only six species. Starting the union operations according to this tree, we have at level 0 the union of all sets (Fig. 4F). Following the hierarchy, InteractiVenn then groups all monocotyledons at level 1, because at this level they are in the same branch (Fig. 4E). It is possible to identify unique sequence clusters of monocotyledons relative to Arabidopsis thaliana. Navigating to level 2 (Fig. 4D), the diagram shows two groups between the monocotyledons: the union of O. sativa, B. distachyon and S. bicolor, and the union of M. acuminata and P. dactylifera. At level 3 (Fig. 4C) there is only one group: O. sativa and S. bicolor. When level 4 is reached, the diagram shows each species’ set individually (Fig. 4B).
More than providing a wider set of features to publish results of sets comparisons, InteractiVenn extends the ability to analyze combinations of sets of elements in part or in total, affording additional observations on the interactions between joined sets. We expect that InteractiVenn will be used by biologists in a wide range of tasks and also by other researchers who seek for more flexible control of the examination of set unions in Venn diagrams of objects.
Availability and requirements
The authors thank Mathieu Rouard and Angélique D’Hont for kindly providing the data used in their study . The authors acknowledge the financial support of Fundação de Amparo à Pesquisa do Estado de São Paulo and of Conselho Nacional de Desenvolvimento Científico e Tecnológico.
- Ruskey F, Weston M. A survey of Venn diagrams. Electron J Comb. 1997; 4.Google Scholar
- D’Hont A, Denoeud F, Aury JM, et al. The banana (musa acuminata) genome and the evolution of monocotyledonous plants. Nature. 2012; 7410:213–7.View ArticleGoogle Scholar
- Hulsen T, de Vlieg J, Alkema W. Biovenn - a web application for the comparison and visualization of biological lists using area-proportional Venn diagrams. BMC Genomics. 2008; 9:488.View ArticlePubMedPubMed CentralGoogle Scholar
- Pirooznia M, Nagarajan V, Deng Y. GeneVenn - a web application for comparing gene lists using Venn diagrams. Bioinformation. 2007; 1(10):420–2.View ArticlePubMedPubMed CentralGoogle Scholar
- Kestler HA, Muller A, Gress TM, Buchholz M. Generalized Venn diagrams: a new method of visualizing complex genetic set relations. Bioinformatics. 2005; 21(5):1592–5.View ArticlePubMedGoogle Scholar
- Bardou P, Mariette J, Escudié F, Djemiel C, Klopp C. jvenn : an interactive Venn diagram viewer. BMC Bioinformatics. 2014; 15(293):1–7.Google Scholar
- Martin B, Chadwick W, Yi T, et al. VENNTURE - a novel Venn diagram investigational tool for multiple pharmacological dataset analysis. PLos One. 2012; 7(5):1–17.View ArticleGoogle Scholar
- Kim Y, Ignatchenko V, Yao CQ, et al. Identification of differentially expressed proteins in direct expressed prostatic secretions of men with organ-confined versus extracapsular prostate cancer. Mol Cell Proteomics. 2012; 11(12):1870–84.View ArticlePubMedPubMed CentralGoogle Scholar
- Kawahara R, Meirelles GV, Heberle H, et al. Integrative Analysis to Select Cancer Candidate Biomarkers to Targeted Validation. In revision.Google Scholar
- Pham T, Piersma S, Warmoes M, Jimenez C. On the beta-binomial model for analysis of spectral count data in label-free tandem mass spectrometry-based proteomics. Bioinformatics. 2010; 26(3):363–9.View ArticlePubMedGoogle Scholar
- Tibshirani R, Hastie T, Narasimhan B, Chu G. Diagnosis of multiple cancer types by shrunken centroids of gene expression. PNAS. 2002; 99(10):6567–572.View ArticlePubMedPubMed CentralGoogle Scholar
- Guyon I, Weston J, Barnhill S, Vapnik V. A gene selection method for cancer classification using support vector machines. Mach Learn. 2002; 46(1-3):389–422.View ArticleGoogle Scholar
- Christin C, Hoefsloot HCJ, Smilde AK, et al. A critical assessment of feature selection methods for biomarker discovery in clinical proteomics. Mol Cell Proteomics. 2013; 12(1):263–76.View ArticlePubMedGoogle Scholar
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.