CSC8309 -- Gene Expression and Proteomics

Analysing the experiment

You should now have familiarised yourself with the basic workings of Cytoscape, and with some ways of adding value to your data using it. We will now apply these techniques, and a few others, to the data you were given at the start of the practical.

Install the APID2NET plugin

APID2NET can be used to draw the protein-protein interaction network of a protein or set of proteins. This can be useful in pulling together more information about the set of proteins identified in a proteomics experiment, particularly when the list is relatively short.

Start a new Cytoscape Session. Start APID2NET (Plugins -> APID2NET). APID Retrieval -> Search List. Set the search parameters: Uncheck 'Search InterSpecies Proteins'. Leave 'Search Hypothetical Proteins' checked. Leave 'iPfam interaction validation' unchecked. Set 'Conexion Levels' to 1. Leave 'Experimental methods' at 1.

What would increasing the Experimental Methods number do to the resulting network?

Click OK. Paste the list of UniProt identifiers (one per line) into the 'Lines to Find' box. Make sure 'Find By' is set to 'UNIPROT_ID'. Click 'Find All'. One of the proteins is not found in the APID2NET dataset, this is an unfortunate, but typical situation. Click 'Paint' to draw the network.

What assumptions should you avoid when considering an APID network?

APID provides a lot of information about individual nodes and edges in the networks it generates. If you select a node in an APID network, and click on the 'APID Info Nodes' button in the Data Panel, you can bring up a panel of data including information from the UniProt entry, InterPro Domains, GO terms etc. The plugin also provides methods of searching the network for particular information, such as GO terms, PFam domains and network hubs (the most highly connected nodes).

On Cytopanel 1, find the 'Node GO' tab and select it. Click on the 'Find GO' button to find all the Biological Process terms in the network (they will be ordered by the number of times they occur). You can then click on individual terms to highlight the nodes annotated with it in the network, and use the 'Mark' button to mark them permanently, should you wish. Use the BiNGO plugin to examine GO Biological Process over-representation in the larger APID network you have generated.

How do the 2 lists of terms compare? Are there any abundant terms that are not over-represented?

The BiNGO results are quite hard to wade through, because a large network is generated with a lot of more general terms like 'Metabolism' high up the list, but some clues as to the intended aims of the experiment do begin to emerge here.

Use MCODE to find subnetworks of the large APID network. Try BiNGO out on the highest scoring subnetwork. Look through the APID Node info for the top 2 subnetworks. Are there any clues to the original experiment here? Select the 'Mark NODE' tab in cytopanel 1. Click the 'Find HUBS' button. This finds the most highly connected nodes in the network, they are central to how the network is put together. You can select them all by clicking on the line that appears after the Hubs have been found that says something like 'HUBS:x% nodes cont y% edges'. Now you can look at the features represented in these selected nodes, using the 'NODE GO', 'NODE PFam' etc tabs. Use BiNGO to examine statistical over-representation of GO Biological Process terms among just the protein list. Same process as above, but paste genes in from text, rather than selecting from the current network. The results of this analysis should provide the last clues you need to make a reasoned guess as to the original aims of the experiment that identified the 20 proteins.