How Can I Convert Hmmer Hmm Descriptions To Psi-Blast Pssms?

Question

[Alex Richter]

I'm attempting to generate an rpsblast database from a set of protein cluster alignments. For most of them, I could use blastpgp -B <alignment> to generate the PSSM without problem, but several of the alignments (namely the large ones, or the ones with huge sequences) crash (seg-fault) blastpgp (legacy v2.2.23, if it matters).

However, these alignments can be hmmbuilt into HMMER profiles without complaint. I know there has to be a way to convert the HMMs to psiblast profiles, since does it for CDD with PFAM and TIGRFAM. I'm just hoping it something better than exporting the alignment, then blastpgp -- particularly since that's what's crashing.

Anyone got a script to convert them directly?

[Michael Schubert]

Was the script working or were you able to create your own conversion script? If so, please share your findings because I'm looking into the exact same thing right now.

[Niek De Klein]

Can someone explain the script of Yun He? I've never used Perl and I don't understand where I have to put what. If I change the

&read_hmmprof("profile_in_HMMER_format.prof", \@hmmprof);

to my own hmm profile it only prints the effective alphabet and nothing else.

Answer 1

[biobot 0.0.77.a.1099]

This post by Yun He from 2005 on the bioinformatics.org Sequence Search mailing list offers a method via hmmconvert and a script. I haven't tried it myself, nor have I tried to check to see if it's correct.

[Alex Richter]

Looks interesting, although it's missing a couple columns in the PSSM, and I'm not entirely sure where the $your_query_sequence is coming from. I'll have to see if I can get it to work. If nothing else, it'll give me a place to start.