PanDDA2 for XChem

[rowanwalker96]

Auto parallel PanDDA2 pipeline for XChem/OpenBind.

Runs downstream from dimple, and waits until all upstream autoprocessing jobs & related dimple jobs have completed before triggering. The 'best' dataset is taken forward to use for PanDDA processing, based on user defined heuristic (I/sigIcompleteness# unique reflections as default as in XCE).

The pipeline is comprised of two stages:

1. PanDDA2 hit identification and auto ligand fitting. This can not begin until a sufficient number of comparator datasets have been collected. Once the number of datasets collected in the auto_model_building directory is equal to the prerun-threshold parameter, a SLURM array job is launched, allocating 1cpu per job. Any dataset that arrives after this point is launched as a single cpu job.
2. PanDDA2 postrun for collating event and site info. Only one PanDDA2_post job can run at any given time.

A custom Rhofit pipeline is also triggered after the PanDDA process step, for high quality ligand fitting to the PanDDA event maps. This results in a final autobuild of the protein and ligand in a dataset's modelled_structures directory.

The pipeline reads the ligand information for each dataset from the XChem soakDB .sqlite database, generates ligand restraints and auto creates the XChem directory structure, such that XCE can read the status of the XChem experiment via its update datasource from filesystem method.

User options for the autoprocessing can be specified by a .user.yaml file in the labxchem visit directory, but is not a requirement. Currently this has minimal options but will be fleshed out in due course:

data:
  acronym: A71EV2A

autoprocessing:
  pandda:
    prerun-threshold: 300
    heuristic: 'default'
    export: True

  ligand_build: True

Initially will only run on OpenBind visits lb42888 to allow for a testing phase.