Protein Annotation · LiteFold
InterPro
Integrated protein family, domain, and functional-site signatures for sequence classification and annotation.
What InterPro is
InterPro is a database of curated protein signatures that unifies protein family, domain, repeat, and functional-site classifications from multiple member databases into single integrated entries. Published in this format by LiteFold, the tabular resource brings together sequence classification signatures alongside functional and structural annotations.
The dataset addresses the challenge of navigating fragmented signature libraries across different resources. By grouping related signatures and mapping them to Gene Ontology terms, Protein Data Bank structures, taxonomic distributions, and literature references, it provides researchers with an integrated framework to explore protein function and sequence organisation.
How it was made
The dataset compiles InterPro release 108.0 into structured tabular records. Each row corresponds to a single InterPro entry and records its accession identifier, classification class (such as Family, Domain, or Active_site), textual abstract, and hierarchical relationships (parent and child identifiers, as well as tree depth).
Entries integrate cross-references and signatures across 18 member database sources, linked Gene Ontology (GO) terms across three namespaces, associated Protein Data Bank (PDB) structure identifiers, taxonomic distributions, and PubMed citations. The data is partitioned into deterministic training and test splits using a hash of the InterPro accession (sha256 modulo 10).
Key facts
- InterPro release
- 108.0 (29th January 2026)
- Total entries
- 51,489
- Train split entries
- 46,440
- Test split entries
- 5,049
- Data format
- Parquet
- Data modalities
- Tabular, Text
- Entries with GO mappings
- 14,799
- Entries with PDB structures
- 30,172
- Entries with publications
- 38,194
- Member database sources
- 18
What you can use it for
- Protein functional annotation
- Map uncharacterised protein sequences to known functional roles and Gene Ontology terms using matched InterPro entry categories.
- Structural biology cross-referencing
- Filter domain and family entries to identify corresponding experimental Protein Data Bank structures for comparative modelling or structural analyses.
- Hierarchical protein classification
- Analyse parent-child tree depths and relationships across protein families, homologous superfamilies, and repeats.
- Machine learning benchmark evaluation
- Train and evaluate protein sequence annotation or classification models using the predefined deterministic hash split.
Explore the data
The schema, real rows, full-text search and a SQL filter — read in your browser, nothing downloaded.
Explore
Read the real rows without downloading anything
Reading rows…
LiveRead from LiteFold/InterPro at the moment you asked. Nothing is cached or stored — every row above came from that request.
Strengths and limits
Good at
- Integrates signatures from 18 distinct member databases into standardized entries.
- Provides predefined, deterministic train-test splits based on hashed InterPro accession identifiers.
- Contains mapped Gene Ontology terms across molecular function, biological process, and cellular component namespaces for 14,799 entries.
Watch out for
- Post-translational modification (PTM) and binding site entry types are represented by very few rows (17 and 82 entries, respectively).
- Only 14,799 of the 51,489 entries contain mapped Gene Ontology terms.
- Hierarchical tree depth and parent-child links are only available where defined in the source hierarchy files.
Questions
Can I look inside InterPro without downloading it?
Yes. This page reads its schema, real rows and per-column statistics in the browser, with full-text search and a SQL filter. Nothing is downloaded to your machine.
What licence is InterPro under?
Its authors declare other. Read the terms on the original page before reusing the data.
Which InterPro version is captured in this dataset?
The dataset contains data from InterPro release 108.0, dated 29th January 2026.
How are the training and test splits constructed?
Splits are determined deterministically by running sha256 on the InterPro accession modulo 10. Bucket 0 forms the test split, while buckets 1 to 9 form the training split.
What types of protein entries are included?
Entries include Family, Domain, Homologous_superfamily, Conserved_site, Repeat, Active_site, Binding_site, and PTM.
Does the dataset include text descriptions?
Yes, each entry includes normalized text abstracts alongside short and full entry names.
How to cite it
Paysan-Lafosse, T., Blum, M., Chuguransky, S. et al. (2023) 'InterPro in 2022', Nucleic Acids Research, 51(D1), pp. D418–D427. doi: 10.1093/nar/gkac993.
Written from InterPro's own dataset card. Statements citing a figure that does not appear there are removed before publication. Licence and running details come from our registry.