Restraints

[JohannesKarwou]

Description

Applying the possibility of restraints in Transformato

[agrass15268]

Features:

• Possibility of adding automatic restraints via config.yaml (using restraints: auto in the simulation paramters)

If no further options are specified, these are harmonic restraints with k=3, between the COM of ligand carbons to the COM of surrounding ( between 5 and 15 A of ligand) carbon-alphas, or alternatively the extremities to more local binding site carbon-alphas. Equilibrium distance is initial distance.

Options available: Configuration of k-Value, flat-bottom potential, wellsize for said flat-bottom, linear scaling of restraint indtroduction.

• Possibility of adding manual restraints via config.yaml

Full specification of restraint anchors via MDAnalysis selection syntax. Otherwise, options as with automatic restraints.

All restraints are implemented as openMM CustomCentroidBondForce. No CHARMM implementation is provided.

Problems

• Testing suite requires local copy of omm_*.py files
• Currently, the COM of the surrounding sites for the simple restraint is determined by the initial ligand configuration, not just the common core

TODO:

• [x] Substitute local omm_ files
• [x] Determine COM of surrounding binding site from common-core done: added function generate_simple_selection, ensuring identical anchor points

For further implementation details, visit the restraint docpage.

• restraint tests run through clean on my side

@JohannesKarwou: Please run full test suite and push changes from your fork to wiederm (you might have to merge - the wiederm repo is 1 commit ahead, I don't have pull rights and would like to avoid circular merges with independent histories)

As a side note: a more general issue might be that the equilibrium distance is calculated from the input .pdb, without minimisation. This means that the restraints aren't exactly aligned even at step 0, small forces (e-05 kJ/mol max) try to enforce the PDB structure. This is not really a problem with our structures, but for structures with large changes due to energy minimisation, this could be troubling.

[codecov-commenter]

Codecov Report

Merging #62 (a3ab677) into master (852ff5d) will increase coverage by 1.35%. The diff coverage is 87.81%.

:exclamation: Current head a3ab677 differs from pull request most recent head c1cb547. Consider uploading reports for the commit c1cb547 to get more accurate results