User:Matthew Whiteside/Notebook/Fumigatus Microarray/2009/08/12

From OpenWetWare

Jump to: navigation, search
Project name Main project page
Previous entry      

TASK4: Finding Correlated Networks

Description of Task

Using matisse to find correlated subnetworks in the KEGG metabolic network, using the SreA dataset.

Contents


Program Input

for details about dataset & network inputs to matisse, see entry.

Obtaining Subnetworks

I ran matisse, playing with the input parameters (min size, alpha, etc) and obtained 25 subnetworks. Files in folder ~/.../2/matisse_network_clustering/kegg_metabolic_modules/. Each cluster/module has network file (*.gml), gene correlation heatmap (*g.png), sample correlation heatmap (*s.png), average expression graph (*a.png).

Viewing Subnetworks

To examine the functions etc. of the 25 correlated subnetworks, i converted them into sif format, added relevant node & edge attributes and then loaded them into Cytoscape. Files are in ~/.../2/matisse_network_clustering/kegg_metabolic_modules/cytoscape/. Script is ~/.../kegg_network_analysis/matisse_to_cytoscape_module_conversion.pl. For each subnetwork there are 4 files:

  1. *.sif - simple interaction file format subnetwork description
  2. *.noa - Cytoscape node attribute file with enzyme names.
  3. *_nodeURL.noa - Cytoscape node attribute file links to KEGG gene entry page.
  4. *_edgeURL.eda - Cytoscape edge attribute file links to KEGG compound entry page - this is the common metabolite linking enzymes.
  5. *.eda - Cytoscape edge attribute with edge weights - weights based on number of edges (1/num_edges) - so enzymes that metabolize many compounds (non-specific enzyme) have lower edge weightings. I did not really use this attribute - it was just there as reference.

A began a short description of the files in module_notes.xlsx, but soon noticed that there was a large degree of overlap between some modules. I used scripts ~/.../kegg_network_analysis/module_overlap.pl and module_collapse.pl to reduce the sets (see script for reduction rules). The reduced subnetworks are in the .../final directory - these are the modules to examine in future analysis. see module_notes.xlsx for a description of the modules. There were 15 final subnetworks

Subnetwork functions

The subnetworks mostly deal with core metabolic, energy-generating processes (pyruvate metabolism, glutamate metabolism). It is not surprising that they are affected, considering iron is a essential molecule for these processes (it may also be a consequence of the KEGG metabolic network which mainly has enzyme annotations for these functions). Little interesting hypothesis were found. Also anthranilate-related subnetwork data is inconclusive - its hard to track the metabolism of an intermediate in a pathway. It could be diverted into any number of downstream products, so it would be difficult to determine if anthraniliate is directly important in iron sequestration as an iron-reducing agent.

SO THIS LINE OF ANALYSIS IS BEING DROPPED.

Personal tools