I would like to use STRING for functional analysis of proteome data, as it appears to be quite powerful. Unfortunately I'm working with an organism that isn't in STRING's database, nor is there anything particularly closely related on the species list. STRING requires all proteins to be from a single species on their list. Is there a sound method to use my data in STRING, perhaps using gene orthologs from another species? For example, could I BLAST my differentially expressed protein sequences against the genome of a single model organism (e.g. Arabidopsis) and use the top hits as the gene IDs? I've tried using simple gene names as the input, but typically only about half are present in any single model organism I've tried. I'm aware that there are hazards in using orthologs, but I'm mostly interested in higher-order patterns and GO term enrichment.
In essence, I think you've got the right ideas. If you want to achieve higher sensitivity than that offered by BLAST, you should use profile-HMMs to detect your orthologs. Check out this paper, it is a full-on bioinformatics approach to the task Article Building protein-protein interaction networks for Leishmania...
and points to some tools that can help you out.
You can try the strategy used here in one of our papers a few years ago: Article The Biomphalaria glabrata DNA methylation machinery displays...
Search for the methods section or this paragraph RNA-Seq: Gene annotation and association network. This approach may be a starting point or give you ideas to work from.
you may try to find the homologous genes in another organism, or use the domain-domain approach ..based on Pfam
- Badges
- Science topic