Opened 18 months ago

Last modified 18 months ago

#15080 assigned defect

Better handling of multi-character chain IDs for Modeller

Reported by: Elaine Meng Owned by: pett
Priority: moderate Milestone:
Component: Structure Prediction Version:
Keywords: Cc:
Blocked By: Blocking:
Notify when closed: Platform: all
Project: ChimeraX

Description

There is something seriously weird about this structure, and I can reproduce that Modeller gives a failure message, but I don't know which part of the data that it hates. Even if I delete the ligand before modeling missing segments, it still fails, so maybe it's something with the sequence.

open 7b6w format mmcif fromDatabase pdb
delete ligand
seq chain /AAA
ui tool show "Model Loops"
modeller refine 1/AAA:1:internal-missing numModels 2 fast false adjacentFlexible 1 protocol standard

-> fails with

Modeller error output
Traceback (most recent call last):
File "ModellerModelling.py", line 95, in <module>
a.make()
File "/usr/lib64/python3.8/site-packages/modeller/automodel/loopmodel.py", line 42, in make
AutoModel.make(self, exit_stage)
File "/usr/lib64/python3.8/site-packages/modeller/automodel/automodel.py", line 141, in make
self.homcsr(exit_stage)
File "/usr/lib64/python3.8/site-packages/modeller/automodel/automodel.py", line 624, in homcsr
self.check_alignment(aln)
File "/usr/lib64/python3.8/site-packages/modeller/automodel/automodel.py", line 577, in check_alignment
aln.check()
File "/usr/lib64/python3.8/site-packages/modeller/alignment.py", line 213, in check
self.check_structure_structure(io=io)
File "/usr/lib64/python3.8/site-packages/modeller/alignment.py", line 222, in check_structure_structure
return f(self.modpt, io.modpt, self.env.libs.modpt, eqvdst)
_modeller.ModellerError: rdpdb_303E> No atoms were read from the specified input PDB file, since the starting residue number and/or chain id in MODEL_SEGMENT (or the alignment file header) was not found; requested starting position: residue number " FIRST", chain " AAA"; atom file name: ./7b6w_1.pdb

Change History (1)

comment:1 by pett, 18 months ago

The problem is that the chain ID is "AAA", and the Modeller input is PDB format. Our code truncates the chain ID sent via the PDB file, but there are other parts of the input that also need to know about the truncation, and the chain ID should be remapped when it comes back from Modeller -- so some work to do.

Note: See TracTickets for help on using tickets.