LiteFold/PDB-CCD

PDB Chemical Component Dictionary

This dataset contains a viewer-friendly Parquet table derived from the PDB Chemical Component Dictionary file in this repository. Each row is one chemical component from components.cif.gz.

The original source file remains in the repository. The default datasets configuration uses the normalized Parquet files under data/ so that the Hugging Face Dataset Viewer and load_dataset() can read component records directly.

Splits

The split is deterministic by component identifier: sha256(component_id) % 10. Bucket 0 is test; buckets 1 through 9 are train.

Split	Rows
train	45,045
test	5,009
total	50,054

Dataset Statistics

Field	Value
Components	50,054
Released components	49,292
Obsolete components	762
Components with atoms	49,947
Components with bonds	49,921
Components with descriptors	50,052
Components with identifiers	37,058
Components with PCM annotations	613
Latest modified date in source	2026-04-24

Common component type	Rows
NON-POLYMER	26,117
non-polymer	19,493
L-PEPTIDE LINKING	923
L-peptide linking	580
peptide-like	556
D-saccharide	405
DNA LINKING	325
RNA LINKING	223

Usage

Install the Hugging Face Datasets library:

pip install datasets

Load all splits:

from datasets import load_dataset

ds = load_dataset("LiteFold/PDB-CCD")
print(ds)

row = ds["train"][0]
print(row["component_id"], row["name"], row["formula"])

Load one split:

from datasets import load_dataset

train = load_dataset("LiteFold/PDB-CCD", split="train")
test = load_dataset("LiteFold/PDB-CCD", split="test")

Stream rows without downloading the full table first:

from datasets import load_dataset

stream = load_dataset("LiteFold/PDB-CCD", split="train", streaming=True)
for row in stream.take(5):
    print(row["component_id"], row["canonical_smiles"], row["atom_count"])

Filter released non-polymer components with InChIKeys:

from datasets import load_dataset

ds = load_dataset("LiteFold/PDB-CCD", split="train")
released = ds.filter(
    lambda row: row["release_status"] == "REL"
    and row["component_type"] in {"NON-POLYMER", "non-polymer"}
    and row["inchikey"] is not None
)
print(released[0])

Find components modified after a date:

from datasets import load_dataset

ds = load_dataset("LiteFold/PDB-CCD", split="train")
recent = ds.filter(lambda row: row["modified_date"] is not None and row["modified_date"] >= "2026-01-01")
print(recent[0]["component_id"], recent[0]["modified_date"])

Columns

Column	Description
component_id	Chemical component identifier.
name	Component name from _chem_comp.name.
component_type	Raw _chem_comp.type value.
pdbx_type	Raw _chem_comp.pdbx_type value.
formula	Chemical formula.
formula_weight	Formula weight as a float.
formal_charge	Formal charge as an integer.
mon_nstd_parent_comp_id	Parent component ID for non-standard monomers, when present.
one_letter_code	One-letter code, when present.
three_letter_code	Three-letter code, when present.
pdbx_synonyms	Synonym field from _chem_comp.
synonym_names	Synonyms from pdbx_chem_comp_synonyms.
initial_date	Initial component date.
modified_date	Last modified date.
release_status	Release status, such as REL or OBS.
replaced_by	Replacement component ID, when present.
replaces	Component ID replaced by this component, when present.
atom_ids	Atom identifiers from chem_comp_atom.
atom_elements	Element symbols for atom_ids.
atom_charges	Atom charges.
atom_aromatic_flags	Atom aromatic flags.
atom_leaving_flags	Atom leaving-atom flags.
atom_stereo_configs	Atom stereochemistry flags.
atom_count	Number of atoms.
heavy_atom_count	Number of non-hydrogen atoms.
hydrogen_atom_count	Number of hydrogen atoms.
bond_atom_id_1	First atom ID for each bond.
bond_atom_id_2	Second atom ID for each bond.
bond_orders	Bond order values.
bond_aromatic_flags	Bond aromatic flags.
bond_stereo_configs	Bond stereochemistry flags.
bond_count	Number of bonds.
descriptor_types	Descriptor types such as SMILES, SMILES_CANONICAL, InChI, and InChIKey.
descriptors	Descriptor values.
canonical_smiles	First SMILES_CANONICAL descriptor.
smiles	First SMILES descriptor.
inchi	First InChI descriptor.
inchikey	First InChIKey descriptor.
identifier_types	Identifier types from pdbx_chem_comp_identifier.
identifiers	Identifier values.
systematic_names	Identifier values whose type is SYSTEMATIC NAME.
audit_actions	Audit action types.
audit_dates	Audit dates.
related_component_ids	Related component IDs, when present.
pcm_ids	Protein modification annotation IDs, when present.
pcm_modified_residue_ids	Modified residue IDs from PCM annotations.
feature_types	Feature types from pdbx_chem_comp_feature.
feature_values	Feature values from pdbx_chem_comp_feature.
split_bucket	Deterministic split bucket from sha256(component_id) % 10.

Preparation

The normalization script used to create the Parquet files is included at scripts/prepare_pdb_ccd_dataset.py.