datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
ADMET-CYP-cofolding
OpenADMET CYP PDB cofolding structures v1
This is a coordinate-complete, one-row-per-structure release for the current UniProt PDB cross-references of human CYP1A2, CYP2C9, CYP2D6, and CYP3A4. It contains 16,739 rows: 149 aligned experimental ground truths and 16,590 valid aligned cofolded structures.
Every coordinate file is gzip-compressed mmCIF and every row points to exactly one relative coordinate_path. All structures use the documented cyp3a4_1tqn_conserved_scaffold_ca_v1… See the full description on the dataset page: https://huggingface.co/datasets/dargason/ADMET-CYP-cofolding.anpdb-admet-ai
ANPDB + ADMET-AI
African natural products from ANPDB with ADMET-AI predicted pharmacokinetic, toxicity, and drug-likeness properties for virtual screening, cheminformatics.
Original data: https://african-compounds.org/anpdb/
ANPDB paper: https://doi.org/10.1093/nar/gkaf1186
ADMET predictions: https://github.com/swansonk14/admet_ai
ADMET-AI uses Chemprop models trained on TDC ADMET datasets.
admet