For Rnaseq Enrichment Analysis What Are Some Gene Ontology (Go) Alternatives?

Question

[biotech]

Hi,

I've been able to perform GO enrichment but since 40% of the bacterial genome I'm working has no GO term assigned, its possible I'm loosing important results. I'm looking for alternatives:

I've though about Kegg or COG enrichment.

Thanks for you help, Bernardo

Answer 1

[Emily]

The problem is that things tend to be correlated. If there's not GO, there's often not other kinds of annotation either.

You could try to find orthologous genes in other species, where there is GO annotation, and use that for your enrichment.

[pld]

Mapping GO annotations with orthologs is your best bet. I've used COG before with success but again the ortholog mapping is a requirement for unannotated species. I'd stay away from KEGG, it tends to be very generic.

With bacteria it is possible to extract TF binding motifs. Differences in regulation are at least as important as differences in protein function. It can also be used to cross check your functional mapping, it there are major differences in your promoter for a similar protein, you can infer there will be differences in the actual function of that gene. There are many databases for bacterial TFBS.

[biotech]

Thanks Joe. Do you mean it could be possible to perform enrichment with TFBS? I'll need, for each gene, the TFBS.

[biotech]

Thanks for your reply Emily. In my particular case. I had 1645 genes with GOs and 1870 with COGs, that is, 225 more, and considering that some GOs are very general, it may be worth to perform COG enrichment. I've created a new question: Any COG enrichment tools (for non-model organisms)?

[biotech]

How to do this? 'You could try to find orthologous genes in other species, where there is GO annotation, and use that for your enrichment.' I understand what you mean but how to perform this task?

Answer 2

[pld]

For the mapping of GO:

1. Find some list of species that have GO annotations
2. Find orthologs of each gene in your species to the genes in your reference species
3. The GO terms of the reference genes orthologous to a given gene in your species are your annotations for that gene

So if reference gene xyz has terms GO:123, GO:456, any gene orthologous to that in your species probably has those terms. Be sure to use best reciprocal blast hits, not just one-way blast. Here is a decent read on it: http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2612752/. You can make your own tools to do this, or just use Blast2GO.

As for the TFBS mining, there are several approaches. The most basic of which is to compile a list of binding sites from phylogenetically close species, and perform a search using the collected PSFM. There are tons of websites that will have this data, just poke around for them. Shameless plug: check out CollecTF, a project I worked on. You can search -300 to +100bp of the TSS as a simplified model of a bacterial promoter.

As a note, be sure that your species actually has something orthologous to the TF whose sites you're looking for.

This would provide a looser but still interesting functional profile, if you know what regulators are acting on a given gene, you can infer what process that gene is part of. e.g. if a gene is regulated by LexA, it probably plays a role in stress response.