Refinement workflow output contains cns error file, but workflow is not affected

[AnnaKravchenko]

Refinemenet workflow, e.g. [topoaa]>>[emref] leads to error files titled emref_X.cnserr with:

 %NEXTCD-ERR: Symbol not found:
     if ($mol_shape_$ncount 

and

 %NEXTCD-ERR: Symbol not found:
     if ($mol_fix_origin_$ncount 

Otherwise workflow runs as expected: emref_X.pdb are generatds and the next module has no issues.

Module/Workflow/Library affected

All refinment modules, so emref, flexref, mdref.
But only the 1st refinemen module is affected, so if the workflow is [topoaa]>>[flexref]>>[emref],
then .cnserr files will be only in flexref output.

Steps to reproduce the behavior

1. Go to the HADDOCK3-antibody-antigen tutorial data, edit workflow/scoring-antibody-antigen.cfg
2. Change the emscoring module to emref
3. Comment out "pdbs/haddock3-ens-emref-ensemble.pdb” and "pdbs/af2-models.pdb” to speed up the run
4. Run workflow with haddock3 workflow/scoring-antibody-antigen.cfg
5. Go to run-scoring/1_emref, check emref_1.cnserr.gz

Version

30b1f89

Additional context

If [rigidbody] is present before the refinement module, then no .cnserr is generated.
Error is reprodusible for an ensenble, as well as a single pdb

I assume it’s not CNS-related issue, more so how modules pass parameters to cns script. If I’m wrong - the labels should be adjusted accordingly.

[amjjbonvin]

Without commenting the ensembles - step 3 - I am getting another error related to not being able to pair the chains.

The issue comes from the fact in this example we start from a complex (and not the separate molecules).
In the emscoring module the fix_origin parameter is not used while it is used in the emref module.

In emref, the number of components is set to 2 in the generated input file, but mol_fix_origin_X and mol_shape_X are only defined for the first molecule.

Note that the same issue will happen with the mdref module.

[amjjbonvin]

You can make the workflow work if you insert a emscoring step with nemsteps = 0 before the emref module.

Probably some info is missing in the io file in 0_topoaa

[VGPReys]

Indeed, after some investigations, I come to the conclusion:

• the parameters mol_fix_origin_X and mol_shape_X are defined for each input molecule (and not chains)
• the variables $mol_fix_origin_$ncount and $mol_shape_$ncount is searched in the CNS script based on the number of components (chains) and not input molecules.

Because an input molecule can be composed of a complex, only mol_fix_origin_1 will be defined, but CNS will search for mol_fix_origin_2 (or more).

This error is reported because of the regex search of ERR in the CNS output, and therefore can lead to some fear. But the only thing that happens is that we cannot define the second chain as shape or fix its origin if the input molecules are provided as complexes.

To solve it, I see different approaches:

• always write all 20 possible mol_fix_origin_X and mol_shape_X parameters in the CNS input
• try to parse this specific CNS output and do not consider it as an error
• modify the CNS scripts to not loop over the number of components (chains), but loop over number of molecules (can be a new parameter to put in the CNS input script)