| |
Research Projects | |
|
Related Projects:
EGFR network |
ESR network |
|
|
In silico tools to launch protein interaction analysis: creating focused networks around proteins of interest |
|
|
Please note: this page is still under construction. Menues are not yet functional. Sorry for inconvenience! |
In collaboration with:
|
Our CSHL chapter cover |
In 2008, we published a chapter in CSHL Press: Basic Methods in Protein Purification and Analysis. In this book, most of the chapters address wet bench-based approaches to identify and dissect interactions between a protein of interest and its partners. These well-validated approaches provide common points of entry into the study of new proteins of interest to a research group. Within the past five years, it has become possible to take a complementary approach, which is based on the exploitation of the increasingly comprehensive databases available in the post-genomic era. By combining information available in these in silico resources, it is becoming feasible to develop a relatively extensive network reflecting physical and functional interactions for any protein of interest. Most of these resources do not require specialized knowledge of computer programming to be exploited by molecular biologists, while user-friendly programs such as Cytoscape and Osprey allow researchers to generate their own local resources for proteins of interest.
This chapter provides a step-by-step illustration of how to use open-access resources to develop a protein-targeted network that can be used to generate and test hypotheses. provides an overview of the steps to follow in constructing a network. For network construction, our primary tools will be protein-protein interaction (PPI) databases, canonical pathways databases, genetic interactions from model organisms, and microarray studies. As a concrete example, we will use these resources to develop a network around a protein of interest, the pro-metastasis factor NEDD9 As of 2008, approximately 75 published papers cite NEDD9 as a main or peripheral topic of study. This is far fewer than for much studied proteins such as Rb (>4000), BRCA1 (~6000) and ERK1 (~8000), but is typical of many proteins of current biological interest.
Network generation is useful in providing an interpretive context for direct purification experiments. The data in the network does not simply reiterate results that can easily obtained from direct reading of the primary literature, but provides a physical and functional interaction "landscape" that can serve as a valuable hypothesis generator for subsequent work. Finally, although this article emphasizes the ability to generate a network based predominantly on in silico resources, we also discuss how all of the tools and resources can be used with a custom set of "seeds" derived from the application of techniques in Chapters 1-15 of the Basic Methods in Protein Purification and Analysis book.
Here we provide some color images and source files, as well as some supplementary information |
||||||||||||||||||||||||||||||||||||||||||||||||||||
 |
Fig. 1 Flow Chart for Network Construction. Twisted arrows indicate the need in ID conversion. Click on the image for a larger view, or download a pdf file (38 kb). |
||||||||||||||||||||||||||||||||||||||||||||||||||||
 |
Fig. 2 Nodes, Edges, and Linkout functions in Cytoscape You can click on the image for a larger view, but it is suggested that you go directly to the Cytoscape session ... |
||||||||||||||||||||||||||||||||||||||||||||||||||||
 |
Fig. 4 Options in combining data After the user becomes familiar with a visualization tool, data is imported from options including PPI databases, model organism-based functional interaction databases, and co-expression (e.g. microarray) databases, and pathway maps (expert knowledge). In building a resource, both "core" datasets, reflecting proteins linked to the network seed in many databases, and "context" datasets, reflecting proteins more distantly connected to the seed or only found in 1 or 2 databases, are generated. How this data is used depends on the researcher's ultimate goal. In option 1 (left), only data found in multiple orthogonal datasets is selected, for example; this extreme option would be useful when working with a very well-known gene, to confine the list of targeted candidates to a reasonable size for mid-throughput experiments.. In option 2 (center), proteins found in multiple "core" datasets connecting to the seed are selected, and the intersections between the "context" datasets are added, to define an interaction sphere of high value candidates. In option 3, right, all proteins linked to the seed by any of the search criteria are maintained as a resource that can be mined as needed to build context around high value proteins that emerge as linked to the seed. In this case, it may also be used to run the merged dataset through the STRING resource to potentially retrieve additional connections between the genes, based on orthogonal datasets. Click on the image for a larger view, or download a pdf file (92 kb). |
||||||||||||||||||||||||||||||||||||||||||||||||||||
 |
Fig. 5 Numbers of interactions reported in protein interaction databases Values represent statistics reported on each website (as of May, 2008), or in recent database-linked publications. Click on the image for a larger view. Do you REALLY want to spend your ink cartridge on the color bar graph? In this case, you can download a pdf file (32 kb)... |
||||||||||||||||||||||||||||||||||||||||||||||||||||
 |
Fig. 6 Assembly of different PPI datasets in Cytoscape First and second neighbors from HPRD, BIND, and BIOGRID are displayed You can click on the image for a larger view, but it would be more educational to generate this view directly from the Cytoscape session ... |
||||||||||||||||||||||||||||||||||||||||||||||||||||
 |
Fig. 7 STRING search for NEDD9 interactors Figure 7 illustrates a simple network that was generated in a search in which only the parameter "experiments", reflecting protein-protein interaction information is considered. Other screening options ("neighborhood", "coexpression", etc) are avaliable, visit http://string.embl.de/ to see details... Values such as confidence interval allow weighting of recovered results to change certainty of displayed information. Click image for larger view. |
||||||||||||||||||||||||||||||||||||||||||||||||||||
 |
Fig. 8 First neighbor core Highly validated nodes found to interact with NEDD9 in multiple PPI databases are shown. Interactions among proteins within the group are also indicated; note some nodes are characterized by multiple cross-linkages, reflecting participation by group members in a signaling pathway (see Figure 13), while others are not, in the absence of additional information groups. You can click the image for a larger view, but the image was directly generated from the Cytoscape session ... |
||||||||||||||||||||||||||||||||||||||||||||||||||||
| Fig. 9 Focal adhesion pathway map from KEGG Go directly to KEGG Web site http://www.kegg.com for the most updated version of the focal adhesion pathway. Although NEDD9 is not found in this map, several of its high confidence first neighbors (e.g. FAK, SRC family kinases, p130Cas) are present: data is collected from functionally linked proteins. |
|||||||||||||||||||||||||||||||||||||||||||||||||||||
 |
Fig. 10 First neighbor core Highly validated nodes found to interact with NEDD9 in multiple PPI databases are shown. Interactions among proteins within the group are also indicated; note some nodes are characterized by multiple cross-linkages, reflecting participation by group members in a signaling pathway (see Figure 13), while others are not, in the absence of additional information groups. Click image for larger view. |
||||||||||||||||||||||||||||||||||||||||||||||||||||
 |
Fig. 11 The DroID, PIM and FLIGHT tools. Figure illustrates a simple network that was generated in a Drosophila Interactions Database (DroID) search in which the parameter "predictions from human" was excluded. Search was done using IM Browser v4. from Russel Finley lab, and can be also configured with text/table output only (main DroID page); Cytoscape plug-in is also available. Click image for slightly larger view.  FLIGHT tool allows batch retrieval of homologues from other species (human, mouse, worm or yeast) using a list of Drosophila genes. |
||||||||||||||||||||||||||||||||||||||||||||||||||||
 |
Fig. 12a Oncomine resource The results of a search for NEDD9 in the database. Click image for larger view. |
||||||||||||||||||||||||||||||||||||||||||||||||||||
 |
Fig. 12b Cancer Outlier Profile Analysis Projection for NEDD9 in the Oncomine database, using the Ginos Head and Neck tumor study as source. Click image for larger view. |
||||||||||||||||||||||||||||||||||||||||||||||||||||
 |
Fig. 12c Cancer Outlier Profile Analysis use of the profiling function to discover genes with similar expression profiles to NEDD9 expression in the Toruner Head and Neck tumor study. Click image for larger view. |
||||||||||||||||||||||||||||||||||||||||||||||||||||
 |
Fig. 13 Composite NEDD9 network This network contains PPIs, networks, Aceview predictions, and text-mining. This Figure shows one representation of an assembled NEDD9 network. We have included (pale green) cases where interactions predicted by Aceview (using all search criteria except protein-protein interactions) overlap with genes in the NEDD9 "second neighbor" sphere of interactors. Purple (top left) represents proteins involved in integrin-dependent signaling; yellow (center left) represents a cluster of TGFβ signaling effectors, including APC10 (identified from protein complexes). It is also possible to readily display interactions that are known to be common between NEDD9 and its paralog, BCAR1/p130Cas (upper left, pink node; interactions shown in pink). Finally (lower right, orange), a number of additional NEDD9 interactions are predicted by various means affecting a diverse set of cellular processes. Click image for a larger view, or Download a pdf file (38 kb) |
||||||||||||||||||||||||||||||||||||||||||||||||||||
 |
Fig. 2 alternative Nodes, Edges, and Linkout functions in Cytoscape Web only! You will NOT find this image in the published chapter, nor will you be able generate it base on the Cytoscape session. You will leave the official FCCC web page and find it here |
||||||||||||||||||||||||||||||||||||||||||||||||||||
Table 1. List of bioinformatics resources recently used by the authors
Please note: it is slightly extended compared to published in the chapter
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||