CoolFace
Datasetpublic

Aregay01/AlphaFoldDB

AlphaFoldDB Prediction Index AlphaFoldDB is an open database of predicted protein 3D structures with confidence scores, massively expanding structural coverage for known protein sequences. Splits Split Rows Parquet files train 222,017,452 12 test 24,672,064 2 total 246,689,516 14 The split is deterministic: hash(uniprot_accession) % 10 == 0 goes to test; buckets 1 through 9 go to train. Dataset Statistics Metric Value… See the full description on the dataset page: https://huggingface.co/datasets/Aregay01/AlphaFoldDB.

sourceHugging Facecc-by-4.0updated 4mo agoView on Hugging Face
0likes154downloads
Dataset Card

AlphaFoldDB Prediction Index

AlphaFoldDB is an open database of predicted protein 3D structures with confidence scores, massively expanding structural coverage for known protein sequences.

Splits

SplitRowsParquet files
train222,017,45212
test24,672,0642
total246,689,51614

The split is deterministic: hash(uniprot_accession) % 10 == 0 goes to test; buckets 1 through 9 go to train.

Dataset Statistics

MetricValue
Rows246,689,516
Minimum sequence length5
Approximate median sequence length278
Mean sequence length328.55
Maximum sequence length4,186
Rows without parsed fragment number5,619,027

Latest-version distribution:

Latest versionRows
15,271,725
2347,302
6241,070,489

The mirrored download_metadata.json describes 48 bulk archive files: 16 proteome archives, 30 global-health archives, and 2 Swiss-Prot archives.

Load With datasets

python
from datasets import load_dataset

ds = load_dataset("LiteFold/AlphaFoldDB")
print(ds)

row = ds["train"][0]
print(row)

Load one split directly:

python
from datasets import load_dataset

train = load_dataset("LiteFold/AlphaFoldDB", split="train")
test = load_dataset("LiteFold/AlphaFoldDB", split="test")

Stream rows without materializing the full table locally:

python
from datasets import load_dataset

streamed = load_dataset("LiteFold/AlphaFoldDB", split="train", streaming=True)
first_row = next(iter(streamed))

Construct an AlphaFold DB entry URL from a row:

python
entry_url = f"https://alphafold.ebi.ac.uk/entry/{row['alphafold_id']}"

Filter to current v6 entries:

python
from datasets import load_dataset

train = load_dataset("LiteFold/AlphaFoldDB", split="train")
v6_train = train.filter(lambda row: row["latest_version"] == 6)

For large jobs, prefer streaming or process the Parquet files with a columnar engine such as DuckDB, PyArrow, Polars, or Spark.

Columns

ColumnDescription
uniprot_accessionUniProt accession from accession_ids.csv.
alphafold_idAlphaFold DB identifier, for example AF-Q5VSL9-F1.
latest_versionLatest available AlphaFold DB model version for the entry.
first_residue_indexFirst residue index in UniProt numbering.
last_residue_indexLast residue index in UniProt numbering.
sequence_lengthDerived as last_residue_index - first_residue_index + 1.
fragment_numberParsed F<number> suffix from alphafold_id, nullable when the suffix is absent or nonstandard.
is_fragmented_predictionWhether fragment_number is greater than 1.
split_bucketDeterministic bucket from hash(uniprot_accession) % 10; bucket 0 is test.

Citation

@article{varadi2022alphafolddb,
  title   = {{AlphaFold} Protein Structure Database: massively expanding the structural coverage of protein-sequence space with high-accuracy models},
  author  = {Varadi, Mihaly and Anyango, Stephen and Deshpande, Mandar and others},
  journal = {Nucleic Acids Research},
  volume  = {50},
  number  = {D1},
  pages   = {D439--D444},
  year    = {2022},
  doi     = {10.1093/nar/gkab1061}
}