Can I do a local BLASTp with a PFAM domain as query?

Question

[Gerardo]

Hi mates, I was reading a paper which it states they identify protein sequences in Halyomorpha halys genome performing a local BLASTp search using the PFAM domain (PF00112) as query (paper) It is driving me mad because I couldn't do it. The main doubt I have is how can I be able to add a PFAM domain as query in a BLASTp? is it that possible? Because PFAM domains are HMM and not a unique sequence.

The main idea behind it is to search for papain family cysteine protease domains in the proteome of Halyomorpha halys

Thanks in advance,

Answer 1

[Mensur Dlakic]

You are correct that BLASTp can't read HMM files, so it can't be done directly. Either an HMM has to be converted to a single protein sequence, or a search started with a seed alignment that is used to make HMMs.

hmmemit from the HMMer package can output a consensus sequence which would be suitable for BLASTp search. Below is a command for the PF01253 domain and its output:

hmmemit -c PF01253.hmm > PF01253.fas
more PF01253.fas

>PF01253-consensus
AKPKIHIRVEKRNGNKKVTLIRGLEGFGIDLEALAKDLKKKLACGGSVTKDPEKGKEIEI
QGDHRKEVAEFLLEKYGIP

You will most likely get a higher sensitivity by searching with hmmsearch and your HMM of interest against a Halyomorpha halys proteome, than with a converted FASTA sequence and BLASTp. There shouldn't be any major difference in speed either.

[Gerardo]

Thanks a lot Mensur! That's what I was looking for! I'm gonna try that package and see how it goes! Now it is clearer, it's logical to use a consensus sequence which comes from a HMM file!

Regards. Gerardo