Merge pull request #726 from RosettaCommons/jadolfbr/antibody_design
Updates to RAbD and IAM
Rosetta Antibody Design Changes
----------------------------------------------
Add two new antibody design protocols to more evenly sample lengths and clusters from the ab db for design. (even_cluster_mc and even_length_and_cluster_mc). Seem to work very well for analyzing potential binding surface during preliminary design runs.
Make RAbD program fail more gracefully when -s or -l is forgotten.
Change a few names of options to make them more consistent with the rest of the protocol.
-instructions -> -cdr_instructions
-base_instructions -> -base_cdr_instructions
-benchmark_graft_design -> -random_start
Run IAM (InterfaceAnalyzerMover) at end of run on each ensemble by default. Add score information to output score file for each PDB (See below).
Run a dualspace relax on each output PDB and ensemble by default. Run IAM on these as well. Name them special for post-design analysis.
Antibody design enums now have their own manager class.
Bugfixes
------------
Fix a major bug where sequence design fallback strategy was overriding the primary strategy. Add function to test this in unit test.
Fix 'hopping' which can occur for conservative sequence design (design a residue, then use that residue for a possible new set of conservative restypes). Here, we keep track of the native positions and the sequence of the grafted CDR structures/sequences by a separate class (antibody/design/NativeAntibodySeq) that is set in the pose DataCache, which accounts for possible pose-length changes and works with our monte carlo algorithm. Add util functions to help with this. Add unit test for this. Now, we always run conservative design as one would think. No val to arg mutations in L1 stability position ever again. That hydrogen bond you created was just too tempting, wasn't it Rosetta!
Fix deterministic graft-permutations. This is not enabled yet, but will allow you to try all possible CDR combinations. The next step is to integrate this into JD3, where it could be actually useful on a cluster for center member trials...
IAM
-----
Fix up InterfaceAnalyzer to not rely on the main global calculator. Separate logic of IntergroupNeighborsCalculator into its own class, which is called by IAM. This allows IAM to be integrated into any protocol with any number of PDBs. Add a function to IAM to add score information for the pose instead of the Job to be called after analysis. This enables any protocol to use IAM and output score information.
Add output of number of hydrogen bonds to IAM.
Test Changes
------------------
Antibody Designer integration test changes expected.
All integration tests which test or use IAM will change with the addition of an additional score in the output score file.