papyrus

is a Data Source.

Papyrus is a large-scale, standardized and curated dataset of small-molecule bioactivity data aimed at machine learning applications such as QSAR and proteochemometric modeling. It aggregates and normalizes ligand-protein bioactivity measurements (with standardized compound structures, UniProt protein targets, and precomputed molecular and protein descriptors) into quality-graded records, and is distributed as versioned releases. The original release compiled about 60 million data points from ChEMBL and ExCAPE-DB; later versions add further sources and grow the dataset. A high-confidence subset, Papyrus++, is provided alongside the full all-quality dataset. The data are hosted on 4TU.ResearchData and Zenodo, and the companion papyrus-scripts Python package and command-line tool handle downloading, versioning, filtering, searching, and export.

License

CC BY-SA 4.0

Homepage

papyrus

Repository

Unknown

External IDs

No external IDs available

Product Summary

Products

From this Resource
ID Name URL Category Format Description
papyrus.dataset Papyrus Dataset 16896406 Product mixed The Papyrus bioactivity dataset, dist...
papyrus.scripts papyrus-scripts Papyrus-scripts Process python papyrus-scripts, a Python library and...

Details

Papyrus

Overview

Papyrus is a large-scale, standardized and curated dataset of small-molecule bioactivity data, built for machine learning uses such as QSAR and proteochemometric modeling. It normalizes ligand-protein bioactivity measurements into quality-graded records, with standardized compound structures, UniProt protein targets, and precomputed molecular and protein descriptors, and ships as versioned releases.

The original release compiled roughly 60 million data points from ChEMBL and ExCAPE-DB; later versions add more sources and continue to grow. A high-confidence subset, Papyrus++, is provided alongside the full all-quality dataset.

Access

  • The dataset is hosted on 4TU.ResearchData (DOI 10.4121/16896406, CC BY-SA 4.0) and mirrored on Zenodo, as versioned releases.
  • The papyrus-scripts Python package and CLI (MIT, on PyPI and GitHub) download, filter, search, and export the data without loading the full dataset into memory.

There is no interactive web frontend; access is through the dataset files and the scripts.

Upstream sources

ChEMBL and ExCAPE-DB are the core bioactivity sources, and protein targets are mapped to UniProt. ExCAPE-DB itself aggregates ChEMBL and PubChem bioassay data, so PubChem-origin activities enter Papyrus through it.

Citation

Béquignon OJM, Bongers BJ, Jespers W, IJzerman AP, van der Water B, van Westen GJP. Papyrus: a large-scale curated dataset aimed at bioactivity predictions. J Cheminform. 2023;15:3. doi:10.1186/s13321-022-00672-x

Is this information incorrect or incomplete? Request an update.

Created: October 28, 2026 | Last modified: October 28, 2026