Building Regulatory Networks with CyTargetLinker summarizes CyTargetLinker, a Cytoscape app that integrates regulatory interactions into network analysis. CyTargetLinker extracts regulatory interactions from multiple data sources called Regulatory Interaction Networks (RINs) and links them to an initial network. This allows visualization of how transcription factors, microRNAs, drugs, and other regulators interact with existing networks like pathways. CyTargetLinker provides flexible filtering and a user-friendly workflow to help biologists generate and analyze hypotheses about regulatory mechanisms involved in processes like breast cancer and DNA damage response. Future work will focus on supporting additional data sources and statistical outputs to analyze overlapping regulatory interactions.
3. Background
• Our initial use case: biological pathways
o Graphical representation of a biological process
o Intuitive
o Signaling pathways, metabolic pathways
o Contains gene products, proteins, metabolites,
different reactions and interactions
o WikiPathways, Reactome, KEGG, ...
5. Background
• Regulatory elements are often missing in
pathways
o transcription factors
o microRNAs
o drugs
o ...
• Why?
o Is the specific regulator really involved in this
7. Goal
• Integration of regulatory elements in network
analysis
• Using multiple datasources in parallel
• Flexibility and extensibility
• Different regulatory interactions, e.g. TF-
8. Goal
User-friendly for biologists and
bioinformaticians
Development of a Cytoscape app to integrate regulatory interactions into network analysis
10. Where does the data come from?
• Regulatory interactions are loaded from
RINs
o RIN = Regulatory Interaction Networks
o RINs support multiple identifier systems (BridgeDb)
o Existing regulatory interaction databases or user
generated
o Conversion scripts are open source and available on
github
14. CyTargetLinker workflow
Extension of the initial network with microRNA-
target information
predicted and validated miRNA targets from 4 different resources
18. CyTargetLinker workflow
• CyTargetLinker extracts the regulatory
interactions from all RINs relevant for the
initial network
• CyTargetLinker for Cytoscape 3 in pre-
release status (downloadable from website)
• CyTargetLinker for Cytoscape 2.8 can be
installed through the plugin manager
20. microRNAs in breast cancer
199 interactions from miRecords
443 interactions from miRTarBase
Heneghan, H. M., et al. "MicroRNAs as novel biomarkers
for breast cancer."Journal of oncology 2010 (2009).
BCL2 --> target of 4 miRNAs
CDKN1A --> target of 4 miRNAs
RB1 --> target of 2 miRNAs
ERBB2 --> target of 2 miRNAs
...
Identify target genes of specific miRNAs
and start building hypotheses
21. Pathway Extension
ENCODE: the Encyclopedia Of DNA Elements
•Goal: identify all functional elements in the human
genome sequence
•Paper: Architecture of the human regulatory network
derived from ENCODE data --> proximal and distal
transcription factor regulation
•Data is provided as networks
•RINs are provided on the website
24. Next steps
• Conversion of more resources to RINs
• Data retrieval through other services
(webservices, RDF, Neo4j databases)
• Export function to show which interactions
were added in which extension step
• Statistical output - VENN diagram showing
overlapping interactions
25. Summary
• User-friendly Cytoscape app to integrate
regulatory interactions into network analysis
in a flexible and extensible way
• RINs and tutorials can be found on
http://projects.bigcat.unimaas.nl/cytargetlinker
• CyTargetLinker is Open Source:
o http://github.com/mkutmon/cytargetlinker
o http://github.com/mkutmon/rin-creation
Original title "Building Biological Regulatory Networks in Cytoscape using CyTargetLinker" Thanks to the organizers for giving me the opportunity to present our Cytoscape app at the NetBioSIG.
I am first going to tell you a little bit about the background of the project, what were our initial use cases. Then I will walk you through an example to extend a gene network with microRNA information. After that I will present shortly two of the biological applications how CyTargetLinker can be applied in research and finally I will show some of the future development ideas that we want to work on.
As some of you probably know, I am involved in the WikiPathways project, which is one of the online pathway resources which is developed in Maastricht and San Francisco together. So our first use case for CyTargetLinker also started with a pathway. Biological pathways are graphical representations of biological process and they are used in educational books, papers, lab-books or presentations because they are much more intuitive to read and understand. There are different types of pathways, like signaling or metabolic pathways, but mostly the pathways contain gene products, proteins, metabolites and then different reactions and interactions to connect them. There are many different resources online available. As we heard in the previous talk, Reactome is one of the pathway databases, like WikiPathways or KEGG.
So usually we start with a pathway diagram like this, which represents the statin pathway from WikiPathways. You will see all the different elements, like gene products, proteins and metabolites, connected through different interactions, like activation or inhibition, but in most pathways you will not see any regulatory elements like microRNAs or drugs.
We believe that integrating those regulatory elements like transcription factors, microRNAs or drugs are crucial to understand how biological processes work. So why are they not there? Sometimes it is not easy to find out if a specific regulatory interaction is really relevant in one specific pathway in the setting that the pathway describes. Furthermore if we would add all the regulatory elements, the pathway might be way too big and cluttered to put it in a educational book or in a presentation, because you wouldn't see the basic pathway anymore. So we believe that if we look at the pathway in form of a network and would move the problem from a pathway to a network problem, we can provide tools to allow that integration of regulatory elements.
I will shortly present the basic idea of the CyTargetLinker app in Cytoscape. You start with a biological network, that might be a protein-protein network, a metabolic network, a set of unconnected nodes or any other biological network. If you want to integrate microRNA-target gene information, you can find for examples those three interactions in miRecords, which is a database containing validated microRNA target interactions. But it's of course not the only one, so CyTargetLinker also allows to visualize multiple resources at the same time. For example one interaction in miRecords is also present in miRTarBase another validated microRNA resource. And it also adds a new interaction that the user would have missed, just using miRecords. microRNAs are of course not the only regulatory elements. When we look at transcription factors, we can add interactions from the TFe database, the transcription factor encyclopedia. And we want that CyTargetLinker is able to integrate interactions in both direction, so it might be that the original network contains a transcription factor and then the targets will be added in the resulting network. So we want to have integration of mutliple datasource, indicated by the color of the edge, multiple types of regulatory interactions, and all possible directions (either adding only regulators, targets or both).
So in summary, CyTargetLinker should provide a framework to integrate regulatory elements in network analysis. It should be able to use multiple datasources in parallel, it should be flexible and extensible, allowing the user to always use the regulatory interactions he needs for his project. That also means it should not be restricted to any species or regulatory interaction type. And we want to implements this tool as a Cytoscape app, because often the created network will be the initial starting point to identify new hypotheses or perform advanced network analysis with all the different apps that are available for Cytoscape.
So the primarily goal was to develop a Cytoscape app to integrate regulatory interactions that can be used by biologists and bioinformaticians. We have users with basically no network analysis experience that use CyTargetLinker to look for overlapping target genes of microRNAs, to look for possible regulators of a set of genes and so on. But we also have Bioinformaticians as users, because the workflow is very easy and it gives them a lot of freedom in what data to choose.
So you can find the tutorials, source code, and general information on our website (http://projects.bigcat.unimaas.nl/cytargetlinker). CyTargetLinker is available through the plugin manager in Cytoscape 2.8 versions and is currently available as download on our website for Cytoscape 3, but most our figures are already created with the new version, so it will be submitted in the upcoming weeks to the app store and will then be available through the Cytoscape app manager in Cytoscape 3 as well. And now I will walk you shortly through a typical CyTargetLinker workflow, but first I want to introduce where CyTargetLinker gets the regulatory information.
We are providing a set of regulatory interaction networks, shortly called RINs. Those RINs are mostly network representations of online interaction databases. Users can download RINs for different species from the CyTargetLinker website, but we also provide the conversion scripts for all the RINs we provide and more in case we are not allowed to redistribute the data directly because of license issues. Furthermore in the RIN creation step we included BridgeDb - an identifier mapping framework for bioinformatics applications. That means that all RINs, independent from which online database they come from, support the same identifier systems.
So usually the data is downloadable as a Excel file which shows one interaction per line. The identifier systems might be different in each database, so it's not easy to compare them directly. On the website we provide those Excel tables in form of a XGMML file, like this. The network contains microRNAs and gene nodes and the target interactions between them. Since all those networks are provided in XGMML files, they can be visualized and used in Cytoscape directly as well.
http://www.broadinstitute.org/gsea/msigdb/cards/GNF2_MKI67.html ToDo: redo figure with neighborhood connections - so MKI67 in the middle I am not so happy with the red color. It is really hard to read the gene names. And I also don't think it looks good. But that is a matter of taste.
TODO: redo
I think there is a slide after this needed - kind of summarizing the whole workflow.
Extension of 8 known microRNAs involved in breast cancer Extend them with miRecords and miRTarBase - well studied microRNAs - lots of interactions typical cancer related genes show up as targets of multiple microRNAs in this set CDKN1A --> p21
had to do this figure in cytoscape 2.8 because we are not far enough with the wikipathways app in cytoscape 3 but that will be ready soon too .