---
license: apache-2.0
tags:
- biology
- bacteria
- operons
- genomics
- genomes
- proteins
pretty_name: Dataset for identifying operons in bacterial genomes (protein sequences)
size_categories:
- n<1K
---
# Dataset for operon identification in bacteria (Protein sequences)

A dataset of 4,073 operons across 11 bacterial genomes species.

The operon annotations have been extracted from [Operon DB](https://operondb.jp/) and the genome protein sequences have been extracted from [GenBank](https://www.ncbi.nlm.nih.gov/genbank/). Each row contains a set of protein sequences present in the genome, represented 
by a list of protein sequences from different contigs.

We extracted high-confidence (i.e. `known`) operons from Operon DB, filtered out non-contigous operons and only kept genomes with at least 9 known operons. We then matched the operon
annotations with the genomes available in [GenBank](https://www.ncbi.nlm.nih.gov/genbank/) using gene names provided by Operon DB.

## Labels
For each operon, we provide the operon protein indices (`operon_protein_indices` column) which can be used to extract the genes which form an operon. We also include operon names and gene names as metadata.

## Usage
We recommend loading the dataset in a streaming mode to prevent memory errors.
```python
from datasets import load_dataset


ds = load_dataset("macwiatrak/bacbench-operon-identification-protein-sequences", split="test")
item = ds[0]

# fetch operon protein indices
operon_prot_indices = item["operon_protein_indices"]
operon_prot_names = item["operon_protein_names"]

# fetch gene protein sequences of an operon
contig_idx = 0
operon_gene_prot_seqs = [
  item['protein_sequence'][contig_idx][i] for i in operon_prot_indices[contig_idx][0]
]
```

## Split

We recommend using operon identification dataset is a fully unsupervised manner, therefore, the entire dataset makes up a `test` set.

See [github repository](https://github.com/macwiatrak/Bacbench) for details on how to embed the dataset with DNA and protein language models as well as code to evaluate operon identification in a zero-shot fasion. 
For DNA sequence representation of the genome see the [bacbench-operon-identification-dna](https://huggingface.co/datasets/macwiatrak/bacbench-operon-identification-dna)
dataset.

---
dataset_info:
  features:
  - name: taxid
    dtype: string
  - name: strain_name
    dtype: string
  - name: contig_name
    sequence: string
  - name: accession_id
    dtype: string
  - name: gene_name
    sequence:
      sequence: string
  - name: protein_name
    sequence:
      sequence: string
  - name: old_protein_name
    sequence:
      sequence: string
  - name: start
    sequence:
      sequence: int64
  - name: end
    sequence:
      sequence: int64
  - name: strand
    sequence:
      sequence: int64
  - name: protein_sequence
    sequence:
      sequence: string
  - name: operon_protein_names
    sequence:
      sequence:
        sequence: string
  - name: operon_protein_indices
    sequence:
      sequence:
        sequence: int64
  - name: operon_names
    sequence:
      sequence: string
  - name: n_operons
    dtype: int64
  splits:
  - name: test
    num_bytes: 16707131
    num_examples: 11
  download_size: 15258914
  dataset_size: 16707131
configs:
- config_name: default
  data_files:
  - split: test
    path: data/test-*
---