Take MD trajectories from a variety of sources and normalize them.

Each trajectory can be in multiple chunks ("gens" in FAH vocabulary). The processing on each gen will not be duplicated. The files will never be concatenated. Use mdtraj's support for loading multiple files per trajectory when doing your analysis: md.load(["fn1.nc", "fn2.nc"], top="whatever.prmtop")

An extensive cache of metadata is stored in one info.json file per trajectory. Your analysis should start by grepping all the json files, saving the order in which they were grepped, and loading trajectories based on the contents of the json file: md.load(info['ctr']['gens'], top=info['stp']['outtop'])

This package was developed to support combining the output of heterogeneous MD engines (openmm, gromacs, and amber), heterogeneous hpc environments (supercomputers, fah), and slightly heterogeneous topologies (different numbers of waters, lipids, ions). Output trajectories should be immediately ready for any quantitative or visual analysis.

Trajectories are processed in stages. Each stage codename serves as a key in the metadata info.json:

• cnv1: optionally convert trajectories using gmx trjconv -pbc whole
• cnv2: convert to netcdf using mdtraj for use in cpptraj
• stp: strip all but closest N waters, lipids, ions using cpptraj
• ctr: autoimage and center using cpptraj

Other metadata includes:

• raw: input
• meta: project, run, clone information
• path: relevant filesystem paths

• mdtraj >= 1.5 (fix for netcdf writing)
• gmx trjconv from Gromacs
• cpptraj from AmberTools

Currently, you have to edit the source code to customize settings for new projects or project types.