bisectgroup/Leak-CURBER-Km

Leak-CURBER Km P-struct

Summary

Leak-CURBER is a benchmark for enzymatic reaction tasks (NeurIPS 2026). This repository holds the leaderboard split Km P-struct at threshold 0.69. It contains the paper's train and validation rows, merged and shuffled with seed 42. The test set is held out for the leaderboard.

Rows

74,904 rows in one train split.

Split

The P-struct split at threshold 0.69: test proteins have a maximum structural similarity to the training proteins below 0.69 (Foldseek).

Columns

Column	Type	Meaning
ECNumber	large_string	EC number of the measurement (null for some rows).
substrate	large_string	Substrate name.
value	double	Raw value in the unit given in unit.
Organism	large_string	Source organism.
uniprot_id	large_string	UniProt accession of the enzyme.
value_type	large_string	Measured quantity (constant in this file).
unit	large_string	Unit of value (constant in this file).
enzymeType	large_string	wildtype or mutant.
mutations	large_list<element: large_string>	List of mutations (null for wild type).
substrate_id	large_string	Compound id in the source database (null for some rows).
uniprot_match	large_string	How the UniProt id was matched (helper column).
source	large_string	Source database. Merged records use `a
pdbs	large_list<element: large_string>	List of PDB ids (empty or null for many rows).
sequence	large_string	Amino acid sequence of the enzyme (protein input).
smiles	large_string	SMILES of the substrate (molecule input).
log10_value	double	Target: log10 of value.
smiles_key	large_string	Canonical SMILES used for grouping (helper column).
smiles_hash	large_string	SHA-256 of smiles_key (helper column).
uniprot_date	timestamp[ns]	UniProt entry creation date (helper column; used by the Time split).

Target

log10_value, the log10 of the Michaelis constant in mM.

Sources and license

This dataset is released for academic, non-commercial use under CC BY-NC-SA 4.0. The terms of each source database also apply. Users must follow them.

• BRENDA, CC BY 4.0 (https://www.brenda-enzymes.org/license.php).
• SABIO-RK, HITS terms for internal non-commercial research and academic use (https://sabiork.h-its.org/layouts/content/termscondition.gsp). This source restricts use to academic, non-commercial research.
• UniProt, CC BY 4.0 (https://www.uniprot.org/help/license): protein sequences and annotation dates.
• RCSB PDB, CC0 1.0 (https://www.rcsb.org/pages/usage-policy), and AlphaFold DB, CC BY 4.0 (https://alphafold.ebi.ac.uk/faq): structure ids in the pdbs column refer to these databases, where present.

The source column gives the database of each row. A value such as brenda|sabio means the merged row comes from both. Rows per source value: brenda 51,616; sabio 21,382; brenda|sabio 1,906.

Citation

@inproceedings{ahmed2026leakcurber,
  title     = {Leak-CURBER: A Leakage-Controlled Multimodal Evaluation Benchmark for Enzymatic Reaction Tasks},
  author    = {Ahmed, Adhil and Balaji, Roshan and Bhatt, Nirav Pravinbhai},
  booktitle = {Advances in Neural Information Processing Systems},
  year      = {2026}
}

Code: https://github.com/bisect-group/Leak-CURBER

Please also cite the sources this dataset uses:

@article{changBRENDAELIXIRCore2021,
    title = {{BRENDA}, the {ELIXIR} core data resource in 2021: new developments and updates},
    volume = {49},
    issn = {0305-1048},
    url = {https://doi.org/10.1093/nar/gkaa1025},
    doi = {10.1093/nar/gkaa1025},
    number = {D1},
    journal = {Nucleic Acids Research},
    author = {Chang, Antje and Jeske, Lisa and Ulbrich, Sandra and Hofmann, Julia and Koblitz, Julia and Schomburg, Ida and Neumann-Schaal, Meina and Jahn, Dieter and Schomburg, Dietmar},
    month = jan,
    year = {2021},
    pages = {D498--D508},
}

@article{wittigSABIORKUpdatedResource2018,
    title = {{SABIO}-{RK}: an updated resource for manually curated biochemical reaction kinetics},
    volume = {46},
    issn = {0305-1048},
    url = {https://doi.org/10.1093/nar/gkx1065},
    doi = {10.1093/nar/gkx1065},
    number = {D1},
    journal = {Nucleic Acids Research},
    author = {Wittig, Ulrike and Rey, Maja and Weidemann, Andreas and Kania, Renate and Müller, Wolfgang},
    month = jan,
    year = {2018},
    pages = {D656--D660},
}

@article{theuniprotconsortiumUniProtUniversalProtein2025,
    title = {{UniProt}: the {Universal} {Protein} {Knowledgebase} in 2025},
    volume = {53},
    issn = {1362-4962},
    url = {https://doi.org/10.1093/nar/gkae1010},
    doi = {10.1093/nar/gkae1010},
    number = {D1},
    journal = {Nucleic Acids Research},
    author = {{The UniProt Consortium}},
    month = jan,
    year = {2025},
    pages = {D609--D617},
}

@article{armstrongPDBeImprovedFindability2020,
    title = {{PDBe}: improved findability of macromolecular structure data in the {PDB}},
    volume = {48},
    issn = {0305-1048},
    url = {https://doi.org/10.1093/nar/gkz990},
    doi = {10.1093/nar/gkz990},
    number = {D1},
    journal = {Nucleic Acids Research},
    author = {Armstrong, David R and Berrisford, John M and Conroy, Matthew J and Gutmanas, Aleksandras and Anyango, Stephen and Choudhary, Preeti and Clark, Alice R and Dana, Jose M and Deshpande, Mandar and Dunlop, Roisin and Gane, Paul and Gáborová, Romana and Gupta, Deepti and Haslam, Pauline and Koča, Jaroslav and Mak, Lora and Mir, Saqib and Mukhopadhyay, Abhik and Nadzirin, Nurul and Nair, Sreenath and Paysan-Lafosse, Typhaine and Pravda, Lukas and Sehnal, David and Salih, Osman and Smart, Oliver and Tolchard, James and Varadi, Mihaly and Svobodova-Vařeková, Radka and Zaki, Hossam and Kleywegt, Gerard J and Velankar, Sameer},
    month = jan,
    year = {2020},
    pages = {D335--D343},
}

@article{varadiAlphaFoldProteinStructure2024,
    title = {{AlphaFold} {Protein} {Structure} {Database} in 2024: providing structure coverage for over 214 million protein sequences},
    volume = {52},
    issn = {0305-1048},
    url = {https://doi.org/10.1093/nar/gkad1011},
    doi = {10.1093/nar/gkad1011},
    number = {D1},
    journal = {Nucleic Acids Research},
    author = {Varadi, Mihaly and Bertoni, Damian and Magana, Paulyna and Paramval, Urmila and Pidruchna, Ivanna and Radhakrishnan, Malarvizhi and Tsenkov, Maxim and Nair, Sreenath and Mirdita, Milot and Yeo, Jingi and Kovalevskiy, Oleg and Tunyasuvunakool, Kathryn and Laydon, Agata and Žídek, Augustin and Tomlinson, Hamish and Hariharan, Dhavanthi and Abrahamson, Josh and Green, Tim and Jumper, John and Birney, Ewan and Steinegger, Martin and Hassabis, Demis and Velankar, Sameer},
    month = jan,
    year = {2024},
    pages = {D368--D375},
}