CatPred-DB: Enzyme Kinetic Parameters Database

• Paper: CatPred: A comprehensive framework for deep learning in vitro enzyme kinetic parameters
• GitHub: https://github.com/maranasgroup/CatPred-DB

Dataset Description

CatPred-DB contains the benchmark datasets introduced alongside the CatPred deep learning framework for predicting in vitro enzyme kinetic parameters. The datasets cover three key kinetic parameters:

Parameter	Description	Datapoints
kcat	Turnover number	23,197
Km	Michaelis constant	41,174
Ki	Inhibition constant	11,929

These datasets were curated to address the lack of standardized, high-quality benchmarks for enzyme kinetics prediction, with particular attention to coverage of out-of-distribution enzyme sequences.

---

Uses

Direct Use: This dataset is intended for training, evaluating, and benchmarking machine learning models that predict enzyme kinetic parameters from protein sequences or structural features.

Downstream Use: The dataset can be used to train or benchmark other machine learning models for enzyme kinetic parameter prediction, or to reproduce and extend the experiments described in the CatPred publication.

Out-of-Scope Use: This dataset reflects in vitro measurements and may not generalize to in vivo conditions. It should not be used as a sole basis for clinical or industrial enzyme selection without additional experimental validation.

---

Dataset Structure

The repository contains:

• datasets/ – CSV files for kcat, Km, and Ki with train/test splits
• scripts/ – Preprocessing and utility scripts

---

Data Fields

Each entry typically includes:

Field	Description
sequence	Enzyme amino acid sequence
sequence_source	Source of the sequence
uniprot	UniProt identifier
substrate_smiles	Substrate chemical structure in SMILES format
value	Raw measured kinetic parameter value
log10_value	Log10-transformed kinetic value (use this for modeling)
log10km_mean	Log10 mean Km value for the enzyme-substrate pair
temperature	Assay temperature (°C)
ph	Assay pH
ec	Enzyme Commission (EC) number
taxonomy_id	NCBI taxonomy ID of the source organism
group	Train/val/test split assignment
pdbpath	Path to associated PDB structural file (if available)
sequence_40cluster	Sequence cluster ID at 40% identity threshold
sequence_60cluster	Sequence cluster ID at 60% identity threshold
sequence_80cluster	Sequence cluster ID at 80% identity threshold
sequence_99cluster	Sequence cluster ID at 99% identity threshold

---

Source Data

Data was compiled and curated from public biochemical databases, including BRENDA and SABIO-RK, as described in the CatPred publication. Splits were designed to evaluate generalization to enzyme sequences dissimilar to those seen during training. All SMILES were sanitized with RdKit. Broken SMILES were removed.

---

Dataset Splits

Each kinetic parameter (kcat, km, ki) has two split strategies, described below.

Split strategies

Random splits divide the data without regard to sequence similarity. These are useful for a general baseline but tend to overestimate real-world model performance, since training and test enzymes may be closely related.

Sequence-similarity splits (seq_test_sequence_XXcluster) ensure that test set enzymes share less than XX% sequence identity with any enzyme in the training set. This is the more rigorous benchmark — a model that performs well here is genuinely generalizing to novel enzymes rather than recognizing similar sequences it has effectively seen before.

Five strictness levels are provided:

Cluster threshold	Test set stringency
20%	Hardest — test enzymes are very dissimilar to training data
40%	Hard
60%	Moderate
80%	Easy
99%	Easiest — nearly any sequence may appear in test

File Naming

Each split file is named {parameter}-{strategy}_{subset}.csv. The subsets are:

Subset	Contents	When to use
train	Training data only	Model development and hyperparameter tuning
val	Validation data only	Monitoring training, early stopping
test	Test data only	Final benchmark evaluation
trainval	Train + val combined	Retrain final model after hyperparameters are locked in
trainvaltest	All data combined	Train a release model once all evaluation is complete

---

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

Load a subset

>>> from datasets import load_dataset

# Options: "kcat", "km", "ki"
>>> ds = load_dataset("RosettaCommons/CatPred-DB", "kcat")

>>> train = ds["train"]
>>> val   = ds["validation"]
>>> test  = ds["test"]

kcat-random_train.csv: 100%|█████████████████████████████████████████████████████████████████████████████████████████████| 28.0M/28.0M [00:04<00:00, 6.29MB/s]
kcat-random_trainval.csv: 100%|██████████████████████████████████████████████████████████████████████████████████████████| 31.1M/31.1M [00:04<00:00, 6.81MB/s]
kcat-random_trainvaltest.csv: 100%|██████████████████████████████████████████████████████████████████████████████████████| 34.5M/34.5M [00:05<00:00, 6.86MB/s]
Generating train split: 100%|█████████████████████████████████████████████████████████████████████████████████| 18789/18789 [00:00<00:00, 67580.51 examples/s]
Generating validation split: 100%|████████████████████████████████████████████████████████████████████████████| 20877/20877 [00:00<00:00, 78951.22 examples/s]
Generating test split: 100%|██████████████████████████████████████████████████████████████████████████████████| 23197/23197 [00:00<00:00, 74253.91 examples/s]

Key columns

Column	Description
sequence	Enzyme amino acid sequence
uniprot	UniProt identifier
reactant_smiles	Substrate SMILES
value	Raw kinetic value
log10_value	Log₁₀-transformed value — use this as your target
temperature	Assay temperature (°C), nullable
ph	Assay pH, nullable
ec	EC number
sequence_40cluster	Cluster ID at 40% identity — use for similarity-based splits

Recommended split workflow

train + val         →   tune architecture and hyperparameters
trainval + test     →   final benchmark (report results here)
trainvaltest        →   train the final released model on all available data

This three-stage approach is standard practice in ML: you only touch the test set once, and the combined files make it easy to retrain on progressively more data as you move from experimentation to deployment.

Basic training setup

>>> df = ds["train"].to_pandas()

>>> X_seq = df["sequence"]
>>> X_sub = df["reactant_smiles"]
>>> y     = df["log10_value"]

# Drop rows with missing targets or substrates
>>> mask = y.notna() & X_sub.notna()
>>> df   = df[mask]

---

Citation

If you use this dataset, please cite:

BibTeX:

@article{boorla2025catpred,
  title={CatPred: a comprehensive framework for deep learning in vitro enzyme kinetic parameters},
  author={Boorla, Veda Sheersh and Maranas, Costas D.},
  journal={Nature Communications},
  volume={16},
  pages={2072},
  year={2025},
  doi={10.1038/s41467-025-57215-9}
}

APA: Boorla, V. S., & Maranas, C. D. (2025). CatPred: a comprehensive framework for deep learning in vitro enzyme kinetic parameters. Nature Communications, 16, 2072. https://doi.org/10.1038/s41467-025-57215-9

License

MIT - see LICENSE

---

Dataset Card Authors

Jessica Lin, Kuniko Hunter, Manasa Yadavalli, McGuire Metts