DLProteinFormats

Installation

using Pkg
pkg"registry add https://github.com/MurrellGroup/MurrellGroupRegistry"
pkg"add DLProteinFormats"

Quickstart

using DLProteinFormats

data = DLProteinFormats.load(PDBSimpleFlat500);

flat_chains = data[1];

chains = DLProteinFormats.unflatten(
    flat_chains.locs,
    flat_chains.rots,
    flat_chains.AAs,
    flat_chains.chainids,
    flat_chains.resinds)

DLProteinFormats.writepdb("chains-1.pdb", chains) # view in e.g. chimerax or vscode protein viewer extension

Flat atom PDB dataset

Flatom is a dataset that stores biomolecular structures in a minimal flat atom primitive format, with each atom represented as a 28-byte NamedTuple with the following fields:

• element::Int8: element number
• category::Int8: structure category
  • 1: protein residue
  • 2: nucleic residue
  • 3: other (e.g. hetero)
• chainid::Int16: chain identifier
• resnum::Int32: residue number (can be optionally renumbered using the MMCIF file)
• resname::StaticStrings.StaticString{3}: 3-character alphanumeric residue name
• atomname::StaticStrings.StaticString{4}: 4-character alphanumeric atom name
• coords::StaticArrays.SVector{3,Float32}: 3D coordinates

using DLProteinFormats

# load 169k PDB structures at flat vectors of atoms (~500 million atoms total, 28 bytes per atom)
# once downloaded, takes ~6 seconds to load
structures = DLProteinFormats.load(PDBFlatom169K);

# atoms of the first structure
atoms = structures[1].atoms

# use the `stack(::Function, ...)` method to get a struct of array vibe
elements = stack(atom -> atom.element, atoms)
coords = stack(atom -> atom.coords, atoms)
# etc. etc.

# split by chain
chainids = map(a -> a.chainid, atoms)
chains = [atoms[findall(==(id), chainids)] for id in unique(chainids)]

using DLProteinFormats.Flatom

# write to PDB
write_to_pdb("chains-1.pdb", atoms)

# write to MMCIF
write_to_cif("chains-1.cif", atoms)