Datasets & structures

makeshift.utils bundles two dependency-light fetchers — one for example datasets, one for protein structures — plus a set of constants. Both cache under ~/.makeshift/.

Datasets

from makeshift.utils import datasets

datasets.list_datasets()          # registered dataset names
path = datasets.fetch("SHP2_NSH2_CPMG")   # download, verify, cache, extract

fetch downloads a dataset zip, verifies its checksum, caches it, and extracts it — returning the local path. It's how the CPMG example gets its .ucsf planes. You can also point it at a custom url with an expected sha256.

Structures

fetch_structure returns a local path to a PDB structure, downloading it if needed. It infers the source from the identifier, or you can force it:

from makeshift.utils import fetch_structure

fetch_structure("model.pdb")     # local file — returned as-is
fetch_structure("1WRP")          # 4-char PDB id  -> RCSB
fetch_structure("P0DP23")        # UniProt accession -> AlphaFold DB
fetch_structure("P0DP23", source="afdb", version="v4")

source	Fetches from
"auto" (default)	inferred from the identifier
"file"	local path
"rcsb"	RCSB PDB
"afdb"	AlphaFold DB

detect_source(identifier) exposes the inference on its own if you just want to know where an id would resolve.

These feed directly into HYDRONMR and the rigid-body step of a relaxation profile.

Constants and reference tables

makeshift.utils.constants holds amino-acid code maps, backbone atom lists, labelling keywords, and identifier regexes used throughout the package. The bundled reference tables (random-coil shifts, PANAV distributions, BMRB statistics) are accessible via makeshift.data.tables.

Full API

See the Utilities reference.