toolevalxm/BioSeq-Raw

BioSeq: Biological Sequence Processing Pipeline

📑 Paper    |    🌐 Project Page    |    💾 Released Resources    |    📦 Repo

We release the raw data for our processed BioSeq dataset, adopted from the original dataset from the BigBio team.

The data format for each line in the protein_sequences_filtered.jsonl is as follows:

{
  "sequence_id": <the unique identifier for the sequence>,
  "amino_acid_sequence": <the amino acid sequence string>,
  "organism": <the source organism>,
  "annotations": [
    {
      "type": <annotation type (e.g., domain, motif)>,
      "start": <start position>,
      "end": <end position>
    },
    ...
  ],
  "source": <the source database>,
  "metadata": <additional metadata about this sample>
}

Some sequences have empty annotations due to incomplete database entries or sequences that exceed our length constraints.

*Note: Due to ongoing curation efforts, some sequence annotations may be incomplete. We continue to update the dataset as more annotations become available.

Citation

Please cite this dataset as: Fries, J. A., Weber, L., Seelam, N., Altay, G., Datta, D., Garda, S., ... & Villanova del Moral, A. (2022). BigBIO: A Framework for Data-Centric Biomedical Natural Language Processing. arXiv preprint arXiv:2206.15076.