Network analyses

#Network-analyses

Olink data from PREFECT.

colab: clements@sas.upenn.edu

Based on results from manuscript. Showing two different methods:

  • Using STRING to explore protein-protein interaction networks based on databases/litterature
  • Enrichment analysis with our actual expression data

Protein-protien interaction networks

#Protein-protien-interaction-networks

As illustrated by STRING db, basically compiling various sources of information to illustrate proteins that cluster based on databases/litterature. Does not take into account the actual data from current experiment.

Acute effects

#Acute-effects

Here I made a quick exploration of networks in the protiens that were significantly altered T0-T1.

  • Network edges define confidence level, and the general confidence level is set to 'high'.
  • Active interaction sources are all, but excluded textmining and 'Co-occurence'.
  • Clusters are defined by MCL-clustering with inflation parameter 3.

string_acute_normal_res.png

As expected since the input is rather large (132 mapped proteins) the networks gets busy. But it illustrates a general idea of how the proteins relate.

Sustained effects

#Sustained-effects

This is a similar general exploration of sustained effects (protiens that were significantly altered T0-T2)

  • Network edges here define molecular function, but still general confidence level is set to 'high'.
  • Active interaction sources are identical, all except for textmining and 'Co-occurence'.
  • Clusters are defined by MCL-clustering with inflation parameter 3.

string_sustained_normal_res.png

Again quite busy graphics (116 mapped proteins), and one idea would be to focus on a selected part of the network. Or simply remove nodes w/o edges.

SUMMARY: This was just a quick illustration of what information STRING graphics provide. When including all significant proteins the graphics get busy and hard to interpret. I think some kind of pre-filtering would be necessary. Another use is just to explore relations btw protein (for hypothesis generation in the discussion part etc).

One limitation is that pathways with greater coverage (e.g. the chemokines) gets highlighted but this does not necessarily mean that these pathways are the most important, just that we covered a larger portion of that pathway.

Enrichement analysis

#Enrichement-analysis

This will be a quick exploration of enrichment analysis to explore enriched pathways/ontologies. The way I think of enrichment analysis is that we can have at least two different aims:

  • Explore pathways: If we're just looking for hypothesis generation, accurate statistics are probably less important.
  • Describe system biology: This is quite data-heavy and statistics get's really important whenever we are trying to draw conclusions from data. I think this method is almost always preferential but I am still learning the methods and won't be able to just produce some quick results.

So I did some quick-and-dirty explorations, this can be done in many ways and I think we shouldn't trust the numbers/statistics too much. But perhaps it can be used as ground for reasoning in the discussion part of an article. Usually I do this in R but it's been a while since I did it and had some issues the databases that I didn't solve today. So I pasted the protein lists into a web-based enrichment tool. It didn't map all the ID's automatically so no process was FDR-significant but here's a general idea of what you can get out of such a tool.

Barplot of enriched sets

#Barplot-of-enriched-sets

wg_bar.png

Diagram of biological processes

#Diagram-of-biological-processes

wg_dag.png

These are general terms and I think that partly reflect that only half of the ID's were mapped correctly (I could do this more accurately in R) but also that we have a lot of general effects in the data. Obviously we can extract which proteins are related to each term.

Conclusion

#Conclusion

I just wanted to export a few examples for you to get a general idea of what we could include. Let me know what your thoughts were, and if you have any additional ideas. I only had a few hours this morning to do this so that's why it is far from complete!

/Andreas