I have a single text file containing amino acid sequence of ~6000 proteins in FASTA format. All proteins belong to a single species, but different strains. I want to determine COG (www.ncbi.nlm.nih.gov/COG) of each protein. I can have two approaches here;
1. First one is simple. Run BLASTp of my query sequences against COG database (pogseqs.fa) available at;
ftp://ftp.ncbi.nlm.nih.gov/pub/kristensen/thousandgenomespogs/blastdb/
and then look for the best hit for each query, and then see hit belongs to which POG group. But what e-value I should use? And what criteria to select best hit, either bit-score or the identity?
2. Second approach is complex: install the COG software (COGsoft.201204.tar) on the Linux system from the following link;
ftp://ftp.ncbi.nih.gov/pub/wolf/COGs/COGsoft/
And then follow the instructions provided in the Readme file (attached). To summarize, first run PSI-BLAST of all-against-all, then manipulate the data in acceptable format to COGnitor by using different modules of COG. Every step requires time in this approach.
Need suggestions.
Am I choosing the right database/tool to determine the orthologous group of genes?
You'll need to do a couple of things:
1) One to convert the GI accessions to uniprot IDs
Try using the uniprot mapper (I've never used the website version as I always write my own but this looks fine).
http://www.uniprot.org/help/mapping
2) Map the uniprot IDs to OGs
You should probably stick with the version 3 release of eggNOG for this part 4 doesn't seem to have the actual 'UniprotAC2eggNOG' file :
http://eggnog.embl.de/version_3.0/downloads.html
The first part doesn't require any coding but the second one does. Should be quite straight forward though.
For the 2), some pseudo-code would be to:
1. Read in the uniprot2OG file
2. Create a matrix with 2 columns and n rows where n = unique(c(uniprotOGs, yourUniprot)). These are also the rownames.
3. Then in: TotalMatrix[uniprotOGs, FirstColumn]
Hey Muhammad,
Instead of running everything yourself. An alternative could be to instead use resources whereby COGs, NOGs (and other millions of other OGs there are) have already been done for you and available for download.
Examples include eggNOG from embl which uses a blast-approach for assigning OGs -> This is the second approach you suggested -> I would not recommend running it yourself since it would take AGES.
http://eggnog.embl.de/version_4.0.beta/
Here, you could assign your ~6000 proteins (uniprot accessions?) to the OG's that they have already done.
If you don't like the blast-type OGs, you could find a resource where the OGs have been determined via trees instead -> I have less experience but there are bound to be some.
Dave
Thank you Dave. My proteins have NCBI GI accessions. Does any of the OGs can accept my 6000 NCBI GIs ? eggNOG can take maximum upto 30 records only.
You'll need to do a couple of things:
1) One to convert the GI accessions to uniprot IDs
Try using the uniprot mapper (I've never used the website version as I always write my own but this looks fine).
http://www.uniprot.org/help/mapping
2) Map the uniprot IDs to OGs
You should probably stick with the version 3 release of eggNOG for this part 4 doesn't seem to have the actual 'UniprotAC2eggNOG' file :
http://eggnog.embl.de/version_3.0/downloads.html
The first part doesn't require any coding but the second one does. Should be quite straight forward though.
For the 2), some pseudo-code would be to:
1. Read in the uniprot2OG file
2. Create a matrix with 2 columns and n rows where n = unique(c(uniprotOGs, yourUniprot)). These are also the rownames.
3. Then in: TotalMatrix[uniprotOGs, FirstColumn]
Hi Muhammad,
If no other database gives you such information about already annotated COGs, you can try tools like Proteinortho (https://www.bioinf.uni-leipzig.de/Software/proteinortho/) , OrthoMCL (http://orthomcl.org/orthomcl/) , etc (I am sure there are many more!). OrthoMCL has been used for many eukaryotic genomes, proteinortho for few. Proteinortho can be run by one step command after installation (in which all-against-all BLAST is done at the back-end) whereas OrthoMCL is very complex. You can give them a try.
Hope this helps.
Snehal
Hi Muhammad
I know your topic is a little out-of-date but I got in touch with it today, and I have quite the same problem to solve. I'm not a bioinformatician so I'm not so friendly with such issues, anyway I must face it somehow. I'd like to know how did you achieved your COG analysis and, if you are currently involved in such problems, how do you manage it now. I know MEGAN5 can be a tool for this, and I used it once with BLASTp outputs, but I'm looking for less cpu-consuming paths.
Thanks in advance for your help
Regards
Stefano
Dear Stefano,
Actually I did not included this experiment in my previous study as it was taking much time, eventually delaying my publication. Hopefully, if I came across any way, I will definitely share in this thread.
Thanks.
Using the NCBI Batch Web CD-Search Tool (http://www.ncbi.nlm.nih.gov/Structure/bwrpsb/bwrpsb.cgi) you can search up to 40,000 sequences at once against several databases, including COG database. It takes roughly one day though, but it's done on their servers, and you might send several batches. However, I think it's the old COG database before the recent update (http://nar.oxfordjournals.org/content/43/D1/D261.abstract)
Also scanning yourself is now much simpler with the eggNOG (currently) 4.5 HMM library. Even simpler: use the new eggNOG mapper at http://beta-eggnogdb.embl.de/#/app/emapper
You can use WebMGA (http://weizhong-lab.ucsd.edu/metagenomic-analysis/server/cog/). Just upload your sequences and get result.
- Badges
- Science topic
- Similar topics
- Bioinformatics
- Bioinformatics and Computational Biology