🧬🖥 Protein Dimension DB 🖥🧬

Scientific data lake with PLM embeddings, GO annotations and taxonomy representations for all proteins in Uniprot/Swiss-Prot

Current Release (2)

Proteins are sorted by length. All files contain the same sequence of proteins, to make joins and merge operations easier.

Protein Language Model Embeddings 🔢

Several models are used to create computational descriptions (embeddings) of the Swiss-Prot proteins:

Model 🤖	Vector Length 📏	Download Links (By Pooling Method) 🔗
ElnaggarLab/ankh-base	768	Mean (1.5G), Parti (3.0G), Max (1.4G), STD (1.4G)
ElnaggarLab/ankh-large	1536	Mean (2.9G), Parti (6.0G), Max (2.7G), STD (2.7G)
ElnaggarLab/ankh2-ext2	1536	Mean (2.9G), Parti (6.0G), Max (2.7G), STD (2.8G)
ElnaggarLab/ankh3-large	1536	Mean (2.9G), Parti (6.0G), Max (2.7G), STD (2.7G)
facebook/esm2_t30_150M_UR50D	640	Mean (1.3G), Parti (2.5G), Max (1.2G), STD (1.2G)
facebook/esm2_t33_650M_UR50D	1280	Mean (2.4G), Parti (5.0G), Max (2.3G), STD (2.3G)
biohub/ESMC-300M	960	Mean (1.8G), Parti (3.7G), Max (1.7G), STD (1.7G)
biohub/ESMC-600M	1152	Mean (2.2G), Parti (4.5G), Max (2.1G), STD (2.1G)
Profluent-Bio/E1-150m	768	Mean (1.5G), Parti (3.0G), Max (587M), STD (1.4G)
Profluent-Bio/E1-300m	1024	Mean (2.0G), Parti (4.0G), Max (757M), STD (1.8G)
Profluent-Bio/E1-600m	1280	Mean (2.4G), Parti (5.0G), Max (929M), STD (2.3G)
flair-bio/amplify-120m	640	Mean (1.3G), Parti (2.5G), Max (1.2G), STD (1.2G)
flair-bio/amplify-350m	960	Mean (1.9G), Parti (3.7G), Max (1.8G), STD (1.8G)

Autoencoder Embeddings and One-hot Encodings 🔢

Numerical representations of the NCBI taxon IDs and InterproScan categories of each protein. Instead of the original NCBI taxonomy tree, we use the custom taxonomy created by taxallnomy project, because it attributes the same number of parent taxa (genus, family, order…) to each species ID.

Name	Description	Vector Length 📏	Download Links 🔗
Interpro Autoencoder	Encoding of the top 17000 most common InterproScan categories in SwissProt proteins	32	Embeddings (36M)
TaxID Autoencoder	Encoding of the top 6006 most common Taxonomic IDs in SwissProt proteins	32	Embeddings (5.1M)
emb.taxa_profile_256.parquet	Taxa One-Hot Encoding	256	Encodings (62M)
emb.taxa_profile_128.parquet	Taxa One-Hot Encoding	128	Encodings (33M)

Protein Annotations 📚

Gene Ontology annotations of Swiss-Prot proteins, separated by evidence type. We also have made available a parsed version of the DeepLoc dataset.

Gene ontology evidence code groups included:

exp: EXP, IMP, IGI, IPI, IDA, IEP, HTP, HDA, HMP, HGI, HEP
phylo: IBA, IBD, IKR, IRD
curated: IC, TAS
comp: ISS, ISO, ISA, ISM, IGP, RCA
iea: IEA

Gene ontology annotation columns:

id: Uniprot ID;
exp, phylo, curated, comp, iea: GO ID list of positive annotations;
exp_not, phylo_not, curated_not, comp_not, iea_not: Negative annotations (NOTs in GOA);
derived_not: NOTs derived in the Warwick and Dessimoz (2020) article;

DeepLoc annotation columns (subcellular locations and membrane protein types) from the Ødum et al. (2024) article:

id: Uniprot ID;
Membrane, Cytoplasm, Nucleus, Extracellular, Cell membrane, Mitochondrion, Plastid, Endoplasmic reticulum, Lysosome/Vacuole, Golgi apparatus, Peroxisome, Peripheral, Transmembrane, LipidAnchor, Soluble: True / False values;

Name	Content	Download Links 🔗
go.expanded.tsv.gz	Simplified version of GOA. Columns: Uniprot ID, GO ID, Evidence Code, Taxon ID and Ontology	25M
go.mf.parquet	Molecular Functions	14M
go.bp.parquet	Biological Processes	21M
go.cc.parquet	Cellular Components	9.9M
deeploc.parquet	Subcellular Locations (DeepLoc dataset)	173KB
interpro.tsv	InterproScan categories of SwissProt proteins	27M
taxid.tsv	Taxonomic ID of each protein in SwissProt. Columns: uniprot_id, taxid, lineage	46M

Uniprot/Swiss-Prot 🔬

Name	Content	Download Links 🔗
ids.txt	Uniprot Accession IDs	10K
sequences.swissprot.fasta	Aminoacid sequences of SwissProt proteins	201M

Others

Name	Content	Download Links 🔗
taxallnomy.parquet	Parent TaxonIDs in Taxallnomy for each NCBI taxon ID	306M
taxid.obo	NCBI taxonomy graph in OBO format	2.0M
interpro.obo	Interpro categories graph in OBO format	4.1M

Citation

Please cite the following work:

Bibtext:

@inproceedings{AlvesSobrinho2025ProteinDimensionDB,
  author       = {Pit{\'{a}}goras de Azevedo Alves Sobrinho and Tetsu Sakamoto and Wilfredo Blanco Figuerola},
  title        = {Protein Dimension DB: A Unified Protein Repository for Representation Learning and Functional Analysis},
  booktitle    = {BioInformatics: 21st Brazilian Congress, X-Meeting 2025, João Pessoa, Brazil, June 3–6, 2025, Proceedings},
  series       = {Lecture Notes in Computer Science},
  volume       = {16037},
  year         = {2025},
  editor       = {Marcio Dorn and Fabricio Martins Lopes},
  publisher    = {Springer Cham},
  isbn         = {978-3-032-09335-6},
  eisbn        = {978-3-032-09336-3},
  address      = {Cham, Switzerland}
}

APA reference:

Alves Sobrinho, P. de A., Sakamoto, T., & Blanco Figuerola, W. (2025). Protein Dimension DB: A unified protein repository for representation learning and functional analysis. BioInformatics: 21st Brazilian Congress, X-Meeting 2025, João Pessoa, Brazil, June 3–6, 2025, Proceedings (Lecture Notes in Computer Science, Vol. 16037). Springer Cham.

References

[1] Alex Warwick Vesztrocy and Christophe Dessimoz. “Benchmarking gene ontology function predictions using negative annotations”, Bioinformatics, 36, 2020, i210–i218, doi: 10.1093/bioinformatics/btaa466;

[2] Marius Thrane Ødum, Felix Teufel, Vineet Thumuluri, et al. “DeepLoc 2.1: multi-label membrane protein type prediction using protein language models”, Nucleic Acids Research, Volume 52, Issue W1, 5 July 2024, Pages W215–W220, doi: 10.1093/nar/gkae237;

[3] Tetsu Sakamoto and Miguel Ortega. “Taxallnomy Database”, Laboratório de Biodados, UFMG. URL;

protein_dimension_db

Datasets with embeddings and other representations for all proteins in Uniprot/Swiss-Prot