Rosetta membrane blastpgp/psiblast database problem for lips4 generation

Member Site Forums Rosetta 3 Rosetta 3 – General Rosetta membrane blastpgp/psiblast database problem for lips4 generation

Viewing 1 reply thread
  • Author
    Posts
    • #1987
      Anonymous

        I have run Rosetta membrane successful in the past, but it seems that now the blastpgp database necessary to run the run_lips.pl script is obsolete.
        In the lastest BLAST release, blastpgp has been dropped entirely in favor of psiblast.
        I am then blocked and I cannot generate the lips4 file using:
        run_lips.pl

        Is there a way of the run_lips.pl connecting the psiblast, or does it need to be modified?

      • #10285
        Anonymous

          There is currently an effort to update and modernize the membrane related code in Rosetta, which would include the run_lips script, so updating is something that’s on the list of things to do, although we don’t have it completed currently.

          Here’s a quick outline of the changes which would need to be made:

          The key line of code in run_lips.pl with respect to blast is line 80, which basically invokes the command

          $PSIBLAST -i $seq -d $DB -j 2 -h 0.001 -b5000 -v5000 -o $blastout

          To use the more recent version, it looks like that needs to be changed to

          $PSIBLAST -query $seq -db $DB -num_iterations 2 -evalue 0.001 -num_alignments 5000 -num_descriptions 5000 -o $blastout

          The output format is also almost-but-not-quite the same, so the alignblast.pl script needs to be changed. I’ve attached a patch file which shows the necessary changes. (This can be quickly applied with “patch alignblast.pl < alignblast.patch_.txt" - I'd recommend making a backup copy of the old version before you do.) Note that I haven’t benchmarked these changes – There may be some additional setting/parameter changes you would need to make in order to get equivalent performance from the lips predictions for the psiblast versus blastpgp. Caveat experimentor.

      Viewing 1 reply thread
      • You must be logged in to reply to this topic.