Chemistry Dataset · maomlab
ChAFF
Curated chemistry dataset for machine learning benchmarking on molecular properties and bioassays.
What ChAFF is
ChAFF is a chemistry benchmarking dataset developed by Maom Lab as part of its wider pharmacology suite. It is designed to evaluate machine learning models across filtered and curated molecular properties and bioassay measurements.
The dataset provides structured chemical data spanning distinct subsets, such as absorbance and reactivity. It addresses the requirement for standardised molecular structures paired with bioassay identifiers for chemical property prediction and pharmacological analysis.
How it was made
The resource is structured as individual tabular data files corresponding to specific experimental or physical properties, such as absorbance and reactivity. Each record contains chemical structures in SMILES format alongside metadata including bioassay identifiers (AID), compound identifiers, identifier types, and dataset labels.
The dataset does not provide pre-partitioned training, validation, or testing splits, defaulting to a single unsplit collection.
Key facts
- Data format
- CSV
- Input features
- Type, DatasetName, AID, ID, IDType, SMILES
- Subject area
- Chemistry and bioassays
What you can use it for
- Machine learning property benchmarking
- Benchmarking predictive models on curated molecular property subsets including chemical absorbance and reactivity.
- Bioassay data analysis
- Analysing compound activity and assay responses referenced by assay identifiers (AID).
- Molecular representation learning
- Evaluating graph neural networks and language models using standardised SMILES representations.
Explore the data
No licence declared
The authors have not published a licence, which means no permission to redistribute the data has been granted to anyone. The rows are not served here; read them on the original dataset page.Strengths and limits
Good at
- Provides standardised SMILES representations mapped directly to bioassay identifiers.
- Modular structure allows individual property datasets, such as absorbance or reactivity, to be analysed separately.
Watch out for
- Lacks predefined training, validation, and test splits.
- Requires explicit feature schema configuration when loading to prevent type inference errors.
Questions
Can I look inside ChAFF without downloading it?
Not here: it declares no licence, so no permission to redistribute its rows has been granted. Its description and statistics are still shown.
What licence is ChAFF under?
Its authors have not declared a licence, which means no permission to reuse or redistribute it has been granted.
What information is recorded for each compound?
Each entry contains a SMILES representation of the molecule, an assay ID (AID), an individual compound ID, the ID type, the dataset name, and a record type label.
Are pre-split train and test sets provided?
No. The dataset files are provided without train, validation, or test partitions, defaulting to a single combined split.
Which specific chemical properties are available?
The dataset contains modular property files covering measurements such as absorbance and reactivity.
Written from ChAFF's own dataset card. Statements citing a figure that does not appear there are removed before publication. Licence and running details come from our registry.