Downstream Processing

Following successful completion of the auto-processing pipelines, downstream pipelines are run to produce biologically relevant structural models from the integrated and merged diffraction data (MTZ files). This includes phasing and model building pipelines, which use techniques to solve the phase problem and then try to generate and refine structural models for solving unknown structures. Other downstream pipelines use known protein strucutres to find unmodelled features such as ligands or metal ions in the structure. All of these pipelines require some additional information to be uploaded to ISPyB in order to run, such as the protein sequence, anomalous scatterer or a known protein structure.

The phasing pipelines use strategies to generate phases, which when combined with the diffraction intensities measured experimentally, can be used to build a map of the electron density. Two main phasing strategies are used by the downstream pipelines run at Diamond, which are experimental phasing and molecular replacement.

Experimental phasing uses the phenomonon of anomalous dispersion, caused by heavy atoms (e.g. S, Se, metal ions) in the protein structure, that results in Friedel pairs of diffraction spots, which would otherwise be equivalent, to have different intensities. These anomalous differences can be used to get phase information for the heavy atom substructure and in turn can be used to calculate phases for the entire protein structure. For experimental phasing pipelines to run, a heavy element that is known to be in the structure needs to be uploaded to ISPyB in the "Anomalous Scatterer" field. The phasing pipelines that are run at Diamond are as follows:

• FastEP - Quick analysis to determine if data is suitable for experimental phasing
• BigEP - Runs a set of similar experimental phasing and model building pipelines that include:
  • Crank2 - From CCP4 software suite
  • autoSHARP - From Global Phasing software suite
  • Autobuild - From the Phenix software suite

Molecular replacement uses known similar protein structures to calculate phases. These similar structures can be from proteins or fragments with a similar amino acid sequence or calculated structures generated by models such as AlphaFold or RoseTTAFold. At Diamond, molecular replacement is run automatically via the MrBUMP pipeline. For MrBUMP to run, an amino acid sequenece for the protein needs to be uploaded to ISPyB.

Aside from the phasing pipelines, there are set of downstream pipelines are run at Diamond, which aim to use a known structure of the protein to find things that are not modelled in the structure such as ligands or metal atoms. These pipelines are summarised below:

• Dimple - Difference map pipeline. Uses a known protein structure to build an electon density map and finds unmodelled features.
• Ligand fit - Runs following Dimple and attempts to fit a ligand structure into unmodelled density. Requires a SMILES string to be uploaded for the ligand.
• PanDDA2 - Pan-dataset Density Analysis for ligands in crystallographic fragment screening
• Pipedream - Global phasing pipeline for analysis of ligands in high-throughput fragment screening.