README.md

metadata

language: en
license: cc-by-sa-4.0
tags:
  - biology
  - protein
  - uniclust
  - uniclust30
  - uniref30
  - msa
pretty_name: UniClust30 (UniRef30)
size_categories:
  - 10M<n<100M

UniClust30 (UniRef30)

Complete UniClust30 / UniRef30 dataset (release 2023_02) from the Söding Lab, converted from HH-suite A3M format to sharded Parquet.

Contents

Each row represents a UniRef30 cluster with:

• cluster_id: Cluster identifier (e.g. UniRef30 accession)
• representative_id: UniProt accession of the representative sequence
• sequence: Representative protein sequence
• sequence_length: Length of the representative sequence
• sequence_xxh128: xxHash-128 of the representative sequence
• num_aligned: Number of sequences in the A3M multiple sequence alignment
• a3m: Full A3M-formatted multiple sequence alignment for the cluster
• member_count: Number of cluster members (from mapping file)
• member_ids: List of all member UniProt accessions

Source

• Origin: Uniclust / UniRef30
• Release: 2023_02
• Format: Converted from HH-suite ffindex/ffdata A3M database
• Compression: zstd
• Reference: Mirdita et al., "Uniclust databases of clustered and deeply annotated protein sequences and alignments", Nucleic Acids Res. 2017

Usage

from datasets import load_dataset

ds = load_dataset("ConvergeBio/uniclust30")

License

UniClust30 data is available under CC BY-SA 4.0.