OPLMFold Training Data: wwPDB Protein Chains (2026-10-12 Snapshot) Preprocessed for Protein Structure Prediction (Briney Lab)

Provider
Briney Lab, Scripps Research; structures from the wwPDB
License
CC0-1.0 (source)
Size
108 files
Format
JSON, Parquet (in sampled files)
Stored at
Hugging Face (https://huggingface.co/datasets/brineylab/oplmfold-training-data)
Access
Free and public. No account with the data provider is needed.
Category
Structural Biology
Source
https://github.com/briney/oplm

About this dataset

Structure training data for OPLMFold, the structure-prediction head of the open protein language model OPLM from the Briney Lab at Scripps Research. The first source folder, wwpdb/2026-10-12, holds the protein chains of the wwPDB mmCIF archive as of 12 October 2026: 254,424 entries and 1,080,220 chains (37 GiB) preprocessed into OPLM's flat all-atom representation, with the Chemical Component Dictionary table. Each versioned folder is self-describing, with a README, a manifest.json of provenance, counts and rejected entries, an index.parquet of one row per chain with its split, Parquet shards of one row per entry, held-out target lists and checksums. Splits follow a date cutoff: train is released on or before 2021-09-30, with valid, post_cutoff, CAMEO and CASP14 held-out evaluation targets and pre-cutoff entries that share a sequence with a held-out entry set apart. Use it to train or evaluate protein structure prediction and folding models; load with oplm.fold.data.StructureDataset or read the Parquet directly. wwPDB data are CC0 1.0; cite the wwPDB consortium, Nucleic Acids Res. 47, D520 (2019).

How to load this dataset in Python

from datasets import load_dataset
ds = load_dataset("brineylab/oplmfold-training-data")
print(ds)

Data structure

File types in sampled files: JSON, Parquet

Files

Example file: wwpdb/2026-10-12/index.parquet

Analyze this dataset without downloading it

On Scigantic, OPLMFold Training Data: wwPDB Protein Chains (2026-10-12 Snapshot) Preprocessed for Protein Structure Prediction (Briney Lab) opens in a hosted JupyterLab notebook with the data already mounted, so you can read files where they are stored instead of copying them first. Each notebook includes an AI assistant that can see the mounted files and write and run analysis code. Open the dataset page to launch a notebook, or browse the full catalog.

This page is also available as Markdown and as JSON.

Related Structural Biology datasets