Finding uncharacterized proteins in a specific pathway

Question

[eimearrutherford]

As part of a research assignment we have to use bioinformatics tools to identify a protein that could potentially be involved in the dna damage response pathway, but is not proven to be part of it, and I don't know where to start. Any help would be greatly appreciated.

[WouterDeCoster]

You could use co-expression analysis, and see which genes are co-expressed with genes involved in the dna damage response.

[Jean-Karim Heriche]

In my experience, gene co-expression is a poor predictor of many cellular functions. This can be understood with a little biological knowledge. For a given pathway, you should start by testing (using cross-validation if in a semi-supervised or supervised setting) whether the differential expression data you have can actually find known pathway members and with which accuracy using the same method you're going to use to identify your candidate gene.

[WouterDeCoster]

That's an interesting approach for internal validation. To be honest, I've never used it for identifying cellular functions - but I know it's something that's being used for that purpose. Whether it performs well at it... I have no idea about!

[Farbod]

@WouterDeCouster, Is it possible to extract the genes that has some GO term related to "DNA damage response", or even KEGG related to it (if any exist) and use them as the research database?

[WouterDeCoster]

Not for co-expression analysis. For co-expression analysis you would:

1. Take a suitable RNA-seq dataset (>20 samples in relevant tissue)
2. Use e.g. WGCNA
3. Perform GO/KEGG enrichment analysis on the modules
4. Take the module enriched for DNA damage response (if any!)
5. Look for genes in this module which were not yet known to be involved in DNA damage response
6. Validate in independent dataset

[Farbod]

No, not for "co-expression analysis", but as a beginning. I mean first collect a bunch of gene that are related to that pathway and make them blastable database and then blast (tblastx?) in order to identify some hit with the protein we are looking for.

[WouterDeCoster]

We are not looking for sequence homology, but for a gene which is in the same functional pathway. So I don't see how BLAST would help here?

[Farbod]

sorry, I assumed that @eimearrutherford has an unknown protein sequence and wants to examine its potential role in DNA damage response.

Answer 1

[Jean-Karim Heriche]

This is typically a gene function prediction task. There are several ways to go about it. For some ideas, have a look at my paper here.