SwissSidechain Residues

This dataset contains a viewer-friendly Parquet table derived from the SwissSidechain archives in this repository. Each row is one unique residue entry from the L or D SMI archives, with parsed availability and summary fields for PDB, MOL2, CHARMM topology, GROMACS RTP/HDB, bundle files, and rotamer libraries.

The original ZIP archives remain 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 the residue table directly.

The source library comments state that the SwissSidechain entries are copyright of the Swiss Institute of Bioinformatics, with non-profit use allowed when the content is not modified and the copyright statement is retained. Commercial use requires a license agreement. Check the original SwissSidechain terms before commercial use.

Splits

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

Split	Rows
train	400
test	59
total	459

Dataset Statistics

Field	Value
Unique residue entries	459
L entries	230
D entries	229
Entries with PDB	459
Entries with MOL2	459
Entries with CHARMM topologies	437
Entries with GROMACS RTP/HDB	437
Entries with source bundles	459
Entries with backbone-dependent rotamers	454
Entries with backbone-independent rotamers	453
Backbone-dependent rotamer rows	15,256,136
Backbone-independent rotamer rows	11,144

Usage

Install the Hugging Face Datasets library:

pip install datasets

Load all splits:

from datasets import load_dataset

ds = load_dataset("LiteFold/SwissSidechain")
print(ds)

row = ds["train"][0]
print(row["entry_id"], row["residue_code"], row["smiles"])

Load one split:

from datasets import load_dataset

train = load_dataset("LiteFold/SwissSidechain", split="train")
test = load_dataset("LiteFold/SwissSidechain", split="test")

Stream rows without downloading the full table first:

from datasets import load_dataset

stream = load_dataset("LiteFold/SwissSidechain", split="train", streaming=True)
for row in stream.take(5):
    print(row["entry_id"], row["stereochemistry"], row["pdb_atom_count"])

Filter D residues with complete topology files:

from datasets import load_dataset

ds = load_dataset("LiteFold/SwissSidechain", split="train")
d_topologies = ds.filter(
    lambda row: row["stereochemistry"] == "D"
    and row["has_top"]
    and row["has_rtp"]
    and row["has_hdb"]
)
print(d_topologies[0])

Find entries with rotamer data:

from datasets import load_dataset

ds = load_dataset("LiteFold/SwissSidechain", split="train")
with_rotamers = ds.filter(lambda row: row["bbdep_rotamer_rows"] > 0)
print(with_rotamers[0]["entry_id"], with_rotamers[0]["bbdep_rotamer_rows"])

Load metadata tables directly:

import pandas as pd
from huggingface_hub import hf_hub_download

rotamer_path = hf_hub_download(
    repo_id="LiteFold/SwissSidechain",
    repo_type="dataset",
    filename="metadata/rotamer_library_counts.parquet",
)
rotamers = pd.read_parquet(rotamer_path)
print(rotamers.head())

manifest_path = hf_hub_download(
    repo_id="LiteFold/SwissSidechain",
    repo_type="dataset",
    filename="metadata/archive_manifest.parquet",
)
manifest = pd.read_parquet(manifest_path)
print(manifest)

Columns

Column	Description
entry_id	Stable table ID in the form L:<code> or D:<code>.
residue_code	Residue code from the source file names.
stereochemistry	Source stereochemistry group: L or D.
smiles	SMILES string from the SMI archive.
smiles_reference_file	Filename referenced inside the SMI row.
source_*_path	Source archive path used for each parsed format.
has_smi	Whether an SMI file exists.
has_pdb	Whether a PDB file exists.
has_mol2	Whether a MOL2 file exists.
has_top	Whether a CHARMM .top file exists.
has_rtp	Whether a GROMACS .rtp file exists.
has_hdb	Whether a GROMACS .hdb file exists.
has_bundle	Whether the entry appears in L_sidechain.zip or D_residues.zip.
has_png	Whether the source bundle contains a PNG image.
has_bundle_bbdep_lib	Whether the source bundle contains a per-entry backbone-dependent rotamer ZIP.
has_bundle_bbind_lib	Whether the source bundle contains a per-entry backbone-independent rotamer ZIP.
bundle_file_extensions	File extensions found in the source bundle.
pdb_residue_names	Residue names parsed from PDB records.
pdb_atom_names	Atom names parsed from PDB records.
pdb_elements	Element symbols parsed from PDB records.
pdb_atom_count	Number of PDB atoms.
pdb_heavy_atom_count	Number of non-hydrogen PDB atoms.
pdb_hydrogen_atom_count	Number of hydrogen PDB atoms.
mol2_molecule_name	Molecule name from the MOL2 file.
mol2_molecule_type	MOL2 molecule type.
mol2_charge_type	MOL2 charge type.
mol2_atom_names	Atom names parsed from MOL2.
mol2_atom_types	Atom types parsed from MOL2.
mol2_atom_charges	Partial charges parsed from MOL2.
mol2_atom_count	Number of MOL2 atoms.
mol2_bond_count	Number of MOL2 bonds.
mol2_total_partial_charge	Sum of MOL2 partial charges.
charmm_residue_name	Residue name parsed from CHARMM RESI.
charmm_residue_charge	Residue charge parsed from CHARMM RESI.
charmm_residue_comment	Comment attached to the CHARMM RESI line.
charmm_atom_names	Atom names from CHARMM topology.
charmm_atom_types	Atom types from CHARMM topology.
charmm_atom_charges	Atom charges from CHARMM topology.
charmm_atom_count	Number of CHARMM atoms.
charmm_bond_count	Number of CHARMM bond pairs.
gromacs_residue_name	Residue name from GROMACS RTP.
gromacs_residue_comment	Comment attached to the GROMACS residue header.
gromacs_atom_names	Atom names from GROMACS RTP.
gromacs_atom_types	Atom types from GROMACS RTP.
gromacs_atom_charges	Atom charges from GROMACS RTP.
gromacs_atom_count	Number of GROMACS RTP atoms.
gromacs_bond_count	Number of GROMACS RTP bonds.
gromacs_improper_count	Number of GROMACS RTP impropers.
hdb_rule_count	Number of hydrogen database rules.
hdb_declared_rule_count	Rule count declared in the HDB header.
bbind_rotamer_rows	Rows in the backbone-independent rotamer library for this residue code.
bbdep_rotamer_rows	Rows in the backbone-dependent rotamer library for this residue code.
split_bucket	Deterministic split bucket from sha256(entry_id) % 10.

Preparation

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