Merge pull request #955 from RosettaCommons/vmullig/glysymm_problem
Add unit tests for glycine with the -symmetric_gly_tables option
Tasks:
- [x] Add unit tests for each of the talaris2014 terms, confirming that glycine in a given conformation scores identically to glycine in the mirror-image conformation when the flag is used.
- [x] Debug any issues with scoring.
- [x] Add unit tests confirming that L-amino acid conformations (including glycine) score identically to mirror-image D-amino acid conformations (including glycine).
- [x] Debug any issues with scoring:
- [x] yhh_planarity was not scoring D-tyrosine (fixed)
- [x] dslf_fa13 was not currently scoring D-disulfides properly.
- [x] fa_dun was not scoring terminal D-residues properly.
This corrects fa_dun scoring of D-residues at termini, dslf_fa13 scoring of D-disulfides, and dslf_fa13 scoring of mixed D/L disulfides.
All unit tests pass.
The test server is a bit backed up due to hardware outages. Testing on the Baker lab's development box ("jojo"), I see the following integration test changes:
- vancomycin: Due to corrections to fa_dun term for terminal D-residues, resulting in somewhat different packing and minimization trajectories. Structures are still reasonable, and changes are expected.
- ligand_dock_7cpa: This test often produces tiny numerical differences on repeated runs on jojo. I don't know why, but that's what I'm seeing here. There is no visible difference to the output structure, and the energy difference is in the sixth decimal point with the dihedral constraint.
- dcys_disulfidize: Expected changes. More reasonable mixed D/L disulfides are forming, now, as expected.
- d_workflow: Subtle differences, largely due to packer changes at termini, I think. This is expected, given the correction to fa_dun.
- simple_cycpep_predict_symm_gly: Expected changes due to addition of symmetrization of gly params file when the -symmetric_gly_tables option is used.
- posttranslational_mod_io: D-hydroxyproline (0AZ) in isolation (by definition, at a terminus) scores slightly differently because of the fa_dun correction to D-residue scoring at termini.