ML Application Curated SAbDab

Quickstart Usage

Install HuggingFace Datasets package

Each subset can be loaded into python using the Huggingface datasets library. First, from the command line install the datasets library

$ pip install datasets

Optionally set the cache directory, e.g.

$ HF_HOME=${HOME}/.cache/huggingface/
$ export HF_HOME

then, from within python load the datasets library

>>> import datasets

Load model datasets

To load structures from the entire SAbDab dataset, use datasets.load_dataset(...):

>>> dataset_tag = "train"
>>> dataset_models = datasets.load_dataset(
  path = "ProteinMPNN/SAbDab_raw",
  name = f"{dataset_tag}_models",
  data_dir = f"{dataset_tag}")['train']

and the dataset is loaded as a datasets.arrow_dataset.Dataset

>>> dataset_models
Dataset({
    features: [
      'pdb',
      'Hchain',
      'Lchain',
      'model',
      'antigen_chain',
      'antigen_type',
      'antigen_het_name',
      'antigen_name',
      'short_header',
      'date',
      'compound',
      'organism',
      'heavy_species',
      'light_species',
      'antigen_species',
      'authors',
      'resolution',
      'method',
      'r_free',
      'r_factor',
      'scfv',
      'engineered',
      'heavy_subclass',
      'light_subclass',
      'light_ctype',
      'affinity',
      'delta_g',
      'affinity_method',
      'temperature',
      'pmid'
      'abangle',
      'annotation_H',
      'annotation_L',
      'imgt_H',
      'imgt_L',
      'sequence_raw',
      'sequence_H',
      'sequence_L',
      'structure',
      'structure_chothia',
      'in_nr_set',
      'curated_quality_dataset',
      'split'
      ],
    num_rows: 20701
})

which is a column oriented format that can be accessed directly, converted in to a pandas.DataFrame, or parquet format, e.g.

>>> dataset_models.data.column('pdb')
>>> dataset_models.to_pandas()
>>> dataset_models.to_parquet("dataset.parquet")

Dataset Details

Dataset Description

This dataset contains curated data from the SAbDab as of Mar 4, 2026. The SAbDab is a database of antibody structures including experimental details, antibody nomenclature, affinity data and sequence annotations. Additional datasets include a ‘in_nr_set’ that was assembled by querying SAbDab with Max. sequence identity 90% and ‘curated_quality_dataset’ that was assembled by querying SAbDab with Max. sequence identity 90%, bound structures, resolution below 4.0 Å. We also performed multivariate stratification on "antigen_type", "heavy_species", "method", "scfv", "engineered", "light_ctype", "in_nr_set", "curated_quality_dataset" to achieve an 80/10/10 train/test/validation split

• Acknowledgements: We kindly acknowledge the SAbDab team, RosettaCommons, and the following institutions: University of California, Los Angeles; University of Maryland; University of Oregon; University of Michigan; University of Pennsylvania; and the Wistar Institute.

• License: CC-BY 4.0

Dataset Sources

• Paper: Dunbar, J., Krawczyk, K. et al (2014). Nucleic Acids Res. 42. D1140-D1146

Uses

Screening of antibody-antigen interactions, querying structural features of antibodies, and more

Citation

@article{10.1093/nar/gkt1043, author = {Dunbar, James and Krawczyk, Konrad and Leem, Jinwoo and Baker, Terry and Fuchs, Angelika and Georges, Guy and Shi, Jiye and Deane, Charlotte M.}, title = {SAbDab: the structural antibody database}, journal = {Nucleic Acids Research}, volume = {42}, number = {D1}, pages = {D1140-D1146}, year = {2013}, month = {11}, abstract = {Structural antibody database (SAbDab; http://opig.stats.ox.ac.uk/webapps/sabdab) is an online resource containing all the publicly available antibody structures annotated and presented in a consistent fashion. The data are annotated with several properties including experimental information, gene details, correct heavy and light chain pairings, antigen details and, where available, antibody–antigen binding affinity. The user can select structures, according to these attributes as well as structural properties such as complementarity determining region loop conformation and variable domain orientation. Individual structures, datasets and the complete database can be downloaded.}, issn = {0305-1048}, doi = {10.1093/nar/gkt1043}, url = {https://doi.org/10.1093/nar/gkt1043}, eprint = {https://academic.oup.com/nar/article-pdf/42/D1/D1140/3538157/gkt1043.pdf} }

Dataset Card Authors

Miranda Simpson (miranda13nicoles@gmail.com), Becca Lee (beccalee5@g.ucla.edu), Nathaniel Felbinger (nfelbing@umd.edu), Pratyush Dhal (pdhal@umich.edu), Colby Agostino (colby.agostino@pennmedicine.upenn.edu)