Merge pull request #1093 from RosettaCommons/vmullig/fix_cycpep_issues
Make RamaPrePro work with D-amino acids and noncanonicals
Currently, RamaPrePro segfaults if you try to use it with noncanonicals. This pull request allows it to score canonical D-amino acids (using inverted Ramachandran tables, as the vanilla rama term does) and anything with a BACKBONE_AA line in its params file, and to return zero for any other noncanonical (which is nicer than crashing). It also allows it to work properly with cyclic geometry.
Since my new unit tests also revealed problems with fa_rep, fa_atr, fa_sol, and pro_close with cyclic geometry (cyclic permutations of the same cyclic peptide pose were not scoring identically), I've fixed those problems, too.
Tasks:
- [x] Properly invert phi/psi when scoring a canonical D-residue.
- [x] Add unit tests for this.
- [x] Derivatives too.
- [x] Add unit tests for derivatives with L-residues.
- [x] Add unit tests for derivatives with D-residues.
- [x] Use a canonical residue's rama tables if BACKBONE_AA is specified in its params file.
- [x] Derivatives too.
- [x] Add unit tests for this.
- [x] Check that rama_prepro handles cyclization properly (as rama does).
- [x] Make rama_prepro no longer dependent on seqpos. (That is, have it check what a residue is actually connected to at its UPPER_CONNECT, rather than assuming that it's connected to residue n+1.)
- [x] Update the PeptideBondedEnergyContainer to handle cyclic/nonstandard geometry.
- [x] Add unit tests for this.
- [x] Fix errors in pro_close with cyclization revealed by the above unit test.
- [x] Fix errors in fa_sol with cyclization revealed by the above unit test.
- [x] Fix errors in fa_rep with cyclization revealed by the above unit test.
- [x] Fix errors in fa_atr with cyclization revealed by the above unit test.
- [x] Add unit test for symmetric scoring of cyclic peptides.
- [x] Return 0 as the score for anything noncanonical (derivatives, too) passed to rama_prepro.
- [x] Add support for symmetric rama_prepro tables for glycine if the -symmetric_gly_tables flag is passed.
- [x] Add unit test for this.
- [x] Add unit test for minimization, too.
- [x] Re-enable the unit test that was disabled for speed (REVERT d00db89d78cc97af7eb89d375c2f50e12d9fc98a).
- [x] Take out debug output in RamaPreProEnergy.cc (REVERT 8c84be7ac1431ffbfe0052f7caa62d16e2381a74).
- [x] Disable dump_scored_pdb() lines in cyclic_geometry unit test.
- [x] Document all of this somewhere.
- [x] Check effect of PeptideBondedEnergyContainer refactor on cart_bonded.
- [x] Beauty.
This pull request also:
- [x] Turns on fa_elec minimizer unit tests. (Some malicious soul had commented them out!)
@fdimaio , @hahnbeom : This should not affect anything related to the calibration of this score term, or its scoring of canonical L-amino acids.
@dabaker : Once this is done, it will no longer be necessary to revert back to the old rama score term when using the beta_nov15 scorefunction
The remaining integration test changes are as a result to small changes to what the CartesianBondedEnergy iterates over when ligands are present.