Merge pull request #2923 from RosettaCommons/frankdt/frankdt/smart_sewing
This pull request introduces the new SEWING refactor and its associated applications.
This new refactor introduces a number of changes and new features into SEWING, including but not limited to:
· SEWING can now be run without generating an edge file (formerly known as a score file). This “unhashed” mode is default and is recommended for most helical proteins. It also avoids the onerous memory requirements of generating an edge file.
· Legacy SEWING only stored one possible overlap for each pair of matching helices, limiting the number of possible designs. SEWING is now able to sample all alignments that pass a user-specified minimum overlap residue cutoff, which greatly increases the sampling space.
· Legacy SEWING input files contained large amounts of data which could not be held in memory simultaneously; therefore, SEWING had to read from disk periodically throughout the protocol, which led to significant slowdowns. SEWING now reads the entire file from disk at the start.
· SEWING now recovers and outputs the lowest-scoring structure encountered during the assembly process. Legacy SEWING did not store and recover the lowest-scoring assembly during a Monte Carlo protocol, instead outputting whichever structure was present at the end of the run. Users can return to this old behavior by setting the new recover_lowest Boolean option to false.
· SEWING now performs temperature ramping between user-specified minimum and maximum temperature values.
· Users can choose not to include binding partners in their output structures via the output_partner Boolean option.
· SEWING now reads all input information at the beginning of a run.
· SEWING score terms and their weights can now be customized by users in their RosettaScript XML files using the AssemblyScorers subtag. Previously, all scorers and their weights were hard-coded.
· SEWING is now compatible with ligands. Users can specify which residue in their starting structure is the ligand and can either specify its protein contacts manually or, in the case of metals and covalently bound ligands, those contacts can be detected automatically. All ligand contacts and the ligand itself will be preserved as vital residues during the SEWING run.
· The new LigandBindingAssemblyMover allows users to introduce new contacts to existing ligands during the assembly process using user-provided specifications of the ligand’s binding geometry.
· Both vital residues and ligands can now be specified using residue selectors.
· Vital residue information can now be written into the output PDB file; whichever residues so identified in the starting structure will be listed, renumbered, in the output structure as "VITAL" and can be extracted with a ResiduePDBInfoHasLabel ResidueSelector.
· Several new score terms and requirements have been introduced, including terms that allow users to more closely control their assembly’s interactions
with binding partners as well as scorers and requirements that provide compatibility with ligand molecules.
· In AppendAssemblyMover, users can either allow SEWING to automatically detect the boundaries of segments by DSSP (a new feature) or can specify the segments and their secondary structures manually. Manual setting of pose segment starts and ends now occurs in the mover tag rather than via command-line options.
· AppendAssemblyMover can now be set to simply extend the starting helix without adding more helices via the extend_mode Boolean option.
Note that the new SEWING refactor does not currently support discontinuous SEWING. For those cases, users will still need to use Legacy SEWING.
Integration test change to simple_metric_features expected due to old changes in ModelFeatures (associated with LegacySEWING).