Merge pull request #3484 from RosettaCommons/roccomoretti/benchmark_multicore_parallel
Update Benchmark Multicore HPC driver to better handle multiple distinct jobs
Currently, the Multicore HPC driver can't parallelize across multiple calls to submit_hpc_job(). This PR adds such functionality.
This should aid local testing of benchmark protocols (like scientific tests) in the case where you're submitting a number of distinct small jobs (versus a large number of the same job).
notify author
notify list [rosetta-logs@googlegroups.com]
Merge pull request #3427 from RosettaCommons/roccomoretti/parse_resnum_RID
Use ResidueIndexDescriptions to remove Pose useage in parse_my_tags()
This updates a number of diverse classes which did (Pose-dependent) residue number parsing in parse_my_tag(). Now they create and store ResidueIndexDescriptions instead, deferring the pose-dependent parsing until apply time.
This updates ResidueIndexDescription for output and serialization, and adds a few new ResidueIndexDescriptor subclasses, such as "end of chain" and "last residue in pose". It also makes a pose-free version of parse_resnum() which returns a ResidueIndexDescription.
The bulk is the conversion of various Movers & Filters to use ResidueIndexDescriptions versus in-parse_my_tag pose use. As an added benefit, certain classes which used to take only Pose-numbered parameters can now also take PDB numbered parameters.
I shouldn't have changed any existing functionality.
notify author
notify list [rosetta-logs@googlegroups.com]
Merge pull request #3483 from RosettaCommons/roccomoretti/seq_rec_metric
Add a simple metric which can calculate sequence recovery.
There's several options for how to calculate the sequence recovery.
* Standard sequence recovery, which is calculated from a subset of residues, comparing their seqeunce with that of a specified reference structures (potentially a "native")
* PSSM recovery - This is a pass/fail based approach from DeLuca et al 2011 - Basically, everything with a positive PSSM score is counted as "not a mutation" (No reference structure needed.)
* Average PSSM score for designed residues. (No reference structure needed).
* Delta PSSM score for designed residues (Uses reference structure.)
notify author
notify list [rosetta-logs@googlegroups.com]
Merge pull request #3478 from RosettaCommons/vmullig/update_xcode_2
Update xcode project to remove illegal horizontal dependencies and to add protocols_g.5, protocols_h.5, and devel
**This PR also got the xcode build working again, so go ahead and compile away with xcode!** In the process, I cleaned up a host of efficiency issues (mostly copying objects that should have been referenced), comma mistakes (usually commas substituting for semi-colons), and an actual significant error in the RemodelDesignMover (Boolean vectors' elements cannot be addressed, since they're bits instead of bytes). Note: in places where xcode was complaining about `for( ObjectCOP const it : my_vector )`, I changed this to `for( ObjectCOP it : my_vector )` and not `for( ObjectCOP const &it: my_vector )`, because using references of smart pointers defeats the purpose of the smart pointer. (Making it a nonconst smart pointer providing const access to the object is still secure enough, I think, and removing the `const` lets xcode stop thinking, "Hey -- this should be a reference! An unnecessary instantiation is happening if the thing being instantiated is const!".)
It looks like protocols_h has stuff in it now, so I added that, too. And for some reason, devel was never set to update automatically the way that the protocols and core libraries were, so I set it to update automatically.
notify author
notify list [rosetta-logs@googlegroups.com]
Merge pull request #3393 from RosettaCommons/jadolfbr/glycan_benchmarking
Symmetry and Glycan QOL improvements
General
======
- Made symmetry finally work with Glycans properly
- Keep the `GlycanTreeSet` up-to-date with the symmetric state of the pose.
- Made symmetric minimization work with Branched torsions (thanks @fdimaio for help tracking this down!)
- Added serialization routines to `SymmetricEnergies`
- Added `update_noncannonical_connection` function to `SymmetricConformation`
- Added a version of `get_score_function()` that takes a pose in order to give the correct `ScoreFunction` for non-global symmetry setup.
- Cleaned up `MinMover` to not edit the movemap based on omega in the `inner_minimization` call and instead take a const `MoveMap`
- Make the `MoveMapFactory` give a movemap based on the symmetric state of the pose
Bug Fixes
=======
- Fix `DensityFitMetric` and `DensityFitSelector`
- Update noncannonical connections after the virtual to real transition for a residue
New Classes (with unit tests)
==================
- `PerResidueGlycanLayerMetric` outputs the layer number of each glycan residue selected
- `MasterSubunitSelector` selects the master subunit in a symmetric pose. If the pose is not symmetric, returns all residues.
New options
=========
- Added `-output_only_asymmetric_subunit` option that extracts the asymmetric unit during PoseToStructFileRep conversion. No this does not work for silent files
- Added options to RMSD SimpleMetrics to allow a superposition on the subset of residues being calculated.
notify author
notify list [rosetta-logs@googlegroups.com]
Merge pull request #3084 from RosettaCommons/aleaverfay/jd3_combine_cl_and_jobdef_for_input_structs
In Issue #3034 (#3034), @jadolfbr proposed that the Input subtag of the Job tag not be required and that instead, if it is absent, then the command line would be fallen back upon. This would allow you to write a job definition file with several jobs where you could leave unspecified which structures these jobs would be run on, and instead, could pair them with -s on the command line, e.g. as -s *.pdb.
I think I have that now working and have a few new unit tests to convince me that it is.
So this brings up a potential concern. Currently, the SJQ asks to deallocate Poses after the input node that uses them is done with its jobs. If the same 1ubq.pdb file were given to two different jobs, then it would be loaded, discarded, and loaded again. The SJQ made no attempt to test whether two Poses were effectively the same. (Indeed, just knowing that two Poses are initialized from the same PDB file is insufficient to know they are the same; e.g. it's possible to construct both a centroid and fullatom Pose from 1ubq.pdb and these Poses would be quite different). With the new system, the user probably intends for an input structure to be "the same" between different jobs. So then it that case, it would be somewhat wasteful to load the structure, deallocate it, and then load it again repeatedly. To address this, I'm adding a flag -jd3:load_input_poses_only_once. Note that per-job-options controlling how Poses are loaded could mean that the protocol behaves quite differently when this flag is on the command line.
Vikram K. Mulligan 6 years Merging pull request #3084 resulted in a Mac/clang unit test failure. Andrew, could you check on this?[list]
Andy Watkins 6 years Note that this unit test was likely affected by @lqtza's last merge, which definitely touched the scs code.
Jeliazko Jeliazkov 6 years Definitely due to my changes... I am looking at it. Odd that this didn't pop up during standard testing.
Andrew Leaver-Fay 6 years OK, I think I understand the problem. Jeli's new code makes a decision based on the SCS_BlastResult.identity data member, but the unit test does not initialize this anywhere.
https://github.com/RosettaCommons/main/blob/4a18ae197069a706bce92479b7d551dc9f0805a2/source/test/protocols/antibody/grafting/scs_functor_filtering.cxxtest.hh
Jeli, could you expand the existing unit test suite to cover your new code in some level? From what I can tell, the existing test is not expected to filter anything, so the new code is not exercised beyond the "if" check (which fails).
Jeliazko Jeliazkov 6 years Yep. Testing against identity is not something I had in mind when I wrote the test. Thanks for pointing this out to me.
notify author
notify list [rosetta-logs@googlegroups.com]
Merge pull request #3485 from RosettaCommons/jadolfbr/minor_updates
IAM bugfix and add job_tag option to ProtocolSettingsMetric
This PR has a few minor fixes and changes:
- Fixes IAM for a test case that @aleaverfay had. It also adds this test case as an integration test.
- Fixes scoretype setting in `PerResidueEnergyMetric` and adds the option to `TotalEnergyMetric` (which may now need to be renamed) @teetsf Both are now tested in integration tests.
- Adds a better description of custom_type option and updates RMSDMetric docs. @roccomoretti
- Adds an option, `job_tag` that basically passes a string to the score file. This is really useful for rosetta_scripts_JD3 to tag the job name in the scorefile to make it easier to concat and analyze.
notify author
notify list [rosetta-logs@googlegroups.com]
Merge pull request #653 from RosettaCommons/vmullig/npro_not_a_donor
Removes DONOR property from Npro atom type
The Npro atom type was incorrectly listed as a hydrogen bond donor. This results in its Lennard-Jones radius being shrunk, and bad things happening in protocols that iterate over donors and whatnot. We've had the pull request to correct this open for three years, and have agreed at more than one meeting that it should be merged. Frank confirmed in 2016 that while the change does have a small effect on numerical values of scores, it doesn't alter his scientific benchmarks. It seems that everyone's on board with this being merged, so here we go.
***SCORING TESTS WILL SHOW SMALL NUMERICAL CHANGES*** Virtually all integration test trajectories are likely to change. Prepare for lots of observer e-mails as integration tests change. This is expected.
Rocco Moretti 6 years I don't know if this commit is directly responsible, but it looks like this commit is the first one which causes the mac.clang.integration hotspot_hashing test to consistently hard-fail with a timeout error.
It's probably worth looking into if this change substantially alters the runtime of this test (and if so, why and if there's an easy way to correct it.)
Vikram K. Mulligan 6 years Yes, this change likely resulted in small trajectory changes that are resulting in a big change in runtime (possibly because some filter isn't being satisfied as soon or something like that). The best solution might just be to play with the random seed for the test.[list]
Rocco Moretti 6 years It also killed the linux.clang.performance.protocols_optimization_Minimizer_dfpmin_armijo_nonmonotone performance test -- it's running about a third of what it once was (consistently in future releases, too). Is that just because of the particular structure being tested, or is that more of a reflection in a global slowdown of minimization?
Vikram K. Mulligan 6 years Since the only thing that has changed is a database change (removing DONOR from the Npro atom type, with all other changes being corrections to the expected outputs in unit tests), I don't see how this could have fundamentally changed the minimizer. It must be the particular structure being tested.[list]
Rocco Moretti 6 years Over the weekend I ran a runtime length test across various proteins. On the whole there doesn't seem to be any change pre/post commit, so it does look like it may be just a quirk of the particular structure being tested.
notify author
notify list [rosetta-logs@googlegroups.com]