Merge pull request #916 from RosettaCommons/vmullig/layerselect
Create a LayerSelect ResidueSelector (like LayerDesign)
ResidueSelectors were initially created because of frustrations involving LayerDesign, a TaskOperation used both to define selections of residue (core, boundary, surface, based on burial) and also to configure packer behaviour. However, LayerDesign has never properly been replaced by ResidueSelectors because there is no selector to select based on layer. This pull request is intended to rectify this. It also:
Partially solves the headaches of mixed canonical/noncanonical design. (LayerDesign is a pain for that, since it can only define layers -- not regions based on linear sequence -- and can only turn canonicals off and non-canonicals on).
Sets up the infrastructure for letting the aa_composition score term count residue types in different sub-regions of a pose (a common request from users being layer-based scoring).
Tasks:
- Move the layer selection logic from protocols to core so that it can be called by core classes. (This is useful, fundamental code that really should be in core in any case.)
- Correct dependencies in LayerDesign.
- Correct dependencies in RotamerBoltzmannWeight.
- Correct various other dependencies.
- Give the user control over sidechain neighbour parameters (cone width, cone length, sigmoid falloff, etc.)
- Distance falloff.
- Factor by which overall residue count is divided.
- Angular factor. (Note that this is calculated poorly, currently. This needs to be refactored, albeit in a way that preserves the current functionality.)
- Document these in LayerDesign.
- Implement a LayerSelect ResidueSelector.
- parse_my_tag()
- apply()
- Documentation.
- Integration test.
- Tweak the ReadResfile TaskOperation to accept a ResidueSelector as a mask.
- Documentation.
- Integration test.
- While I'm at it, fix the irritating behaviour of the ReadResfile TaskOperation that has it read the resfile repeatedly, at apply time, rather than at parse time. (I'm sure the Blue Gene people are a bit sick of this.)
- Track down and fix current issues:
- dna_interface_design integration test failure to run.
- sequence_profile_constraints integration test failure to run.