Hi all,
Thanks you very much for such a prompt response! :-)
Working now!
Summary of successfully produced models: Filename molpdf DOPE score GA341 score ---------------------------------------------------------------------- 1aho.B99990001.pdb 424.19452 -5177.44580 1.00000 1aho.B99990002.pdb 408.62314 -5174.39111 1.00000 1aho.B99990003.pdb 378.08508 -5317.75439 1.00000 1aho.B99990004.pdb 362.42450 -5208.25195 1.00000 1aho.B99990005.pdb 398.92511 -5020.92676 1.00000
:-) :-)
On Mon, Apr 12, 2010 at 4:07 AM, Thomas Juettemann juettemann@gmail.comwrote:
> You need to specify which part of the sequence of your template you > are using. Not specifying it causes the error message "Alignment > sequence not found in PDB file...". > You can find more information about the alignment format here: > http://salilab.org/modeller/manual/node445.html#alignmentformat > Also, your template seems to be a NMR structure, you need to replace > the structureX with NMR. > > Try using these sequences in your alignment file: > > P1;1aho > sequence:1aho:FIRST:@ END:::::: > VKDGYIVDDVNCTYFCGRNAYCNEECTKLKGESGYCQWASPYGNACYCYK > LPDHVRTKGPGRCH--* > > P1;2kbhA > NMR:2kbh.pdb:FIRST:@ END:::::: > VKDGYIADDRNCPYFCGRNAYCDGECKKNRAESGYCQWASKYGNACWCYK > LPDDARIMKPGRCNGG* > > > > On Mon, Apr 12, 2010 at 08:08, Daniel Fernandez dfernan@gmail.com wrote: > > Hi, > > Please help me if possible, I have been stuck few days by now :-( > > I am trying a very basic approach to use modeller, but in this case I am > > using my own global alignment from clustalW. I am trying a very basic > > protein, target: 1aho, and only one template 2kbh_A > > ClustalW outputs the alignment in PIR format > > (http://www.bioinformatics.nl/tools/crab_pir.html) but is a different > PIR > > format than the one required by modeller. Actually the main difference > is > > that clustalW does not provide the start residue:start chain code:end > > residue:end chain code: parts. > > I do not know how to get that information and modeller apparently does > not > > know either. Here I copy and paste the clustalW PIR format, the PIR > modeller > > format and the errors I get from the model_simple.log. > > CLUSTALW PIR format - sequencealn.pir > >>P1;1AHO_A|PDBID|CHAIN|SEQUENCE > > VKDGYIVDDVNCTYFCGRNAYCNEECTKLKGESGYCQWASPYGNACYCYK > > LPDHVRTKGPGRCH-- > > * > >>P1;2KBH_A|PDBID|CHAIN|SEQUENCE > > VKDGYIADDRNCPYFCGRNAYCDGECKKNRAESGYCQWASKYGNACWCYK > > LPDDARIMKPGRCNGG > > * > > MODELLER INPUT (does not work, not sure why) > >>P1;1aho > > sequence:1aho: : : : ::: 0.00: 0.0 > > VKDGYIVDDVNCTYFCGRNAYCNEECTKLKGESGYCQWASPYGNACYCYK > > LPDHVRTKGPGRCH--* > >>P1;2kbhA > > structureX:2kbh.pdb: : : : ::: 0.00: 0.0 > > VKDGYIADDRNCPYFCGRNAYCDGECKKNRAESGYCQWASKYGNACWCYK > > LPDDARIMKPGRCNGG* > > ERROR MODELLER GIVES ME: > > get_ran_648E> Alignment sequence not found in PDB file: 1 > 2kbh.pdb > > (You didn't specify the starting and ending residue numbers > > and > > chain IDs in the alignment, so Modeller tried to guess > these > > fr > > the PDB file.) > > Suggestion: put in the residue numbers and chain IDs (see > the > > manual) and run again for more detailed diagnostics. > > You could also try running with allow_alternates=True to > > accept > > alternate one-letter code matches (e.g. B to N, Z to Q). > > Traceback (most recent call last): > > File "hoModeller-clustalaln.py", line 14, in <module> > > a.make() # do the homollogy modelling > > File "/n/sw/modeller-9v7/modlib/modeller/automodel/automodel.py", line > 98, > > self.homcsr(exit_stage) > > File "/n/sw/modeller-9v7/modlib/modeller/automodel/automodel.py", line > > 411, > > aln = self.read_alignment() > > File "/n/sw/modeller-9v7/modlib/modeller/automodel/automodel.py", line > > 401, > > aln.append(file=self.alnfile, > align_codes=self.knowns+[self.sequence]) > > File "/n/sw/modeller-9v7/modlib/modeller/alignment.py", line 79, in > append > > allow_alternates) > > _modeller.SequenceMismatchError: get_ran_648E> Alignment sequence not > found > > i > > I am aware that the error advice me to put the start and ending residues > > numbers but I am still not sure how to solve the issue. > > Any help will be strongly appreciated, I need to use modeller as soon as > > possible to do some progress in my research. > > Best, > > -- > > Daniel F. > > > > Department of Statistics, Harvard University > > 1 Oxford Street, Cambridge, MA 02138 > > > > _______________________________________________ > > modeller_usage mailing list > > modeller_usage@salilab.org > > https://salilab.org/mailman/listinfo/modeller_usage > > > > >