Written and maintained by CASRAI Editorial Board
Last updated
STRING is a database of known and predicted protein-protein interactions, covering both physical binding and functional association. This page is for researchers who need to know what the STRING database contains, who maintains it, how its data can be accessed and reused, and how to cite it. Figures below are as read from the STRING website in November 2026 (version 12.5) and change with each release.
What STRING is
STRING (string-db.org) is a database of protein-protein interactions that includes physical and functional associations. Per its About page, the interactions come from computational prediction, from knowledge transfer between organisms, and from interactions aggregated from other primary databases. A functional association means two proteins contribute to a shared function; it does not by itself mean they bind each other.
Version 12.5, described in the 2025 Nucleic Acids Research update, added a regulatory network built from curated pathway databases plus a fine-tuned language model that reads the literature for interaction type and direction. STRING now offers functional, physical and regulatory networks as separate views, along with improved pathway enrichment (better false-discovery-rate correction, redundancy filtering, new visuals) and downloadable network embeddings.
Who runs it and what it covers
STRING is run jointly by the Swiss Institute of Bioinformatics (SIB), the Novo Nordisk Foundation Center for Protein Research (CPR), the University of Zurich and EMBL Heidelberg. It is designated a Core Data Resource by the Global Biodata Coalition and ELIXIR.
| Measure (as read November 2026, v12.5) | Value |
|---|---|
| Proteins | 59,309,604 |
| Organisms | 12,535 (10,756 bacteria, 1,322 eukaryotes, 457 archaea) |
| Interactions, all confidence levels | 28,464,487,748 |
| Interactions with score of at least 0.700 | 953,496,894 |
| Interactions with score of at least 0.900 | 299,233,302 |
The counts show why the confidence score matters: the thresholds of 0.700 and 0.900 keep roughly 3 percent and 1 percent of the stored interactions, so the threshold you choose strongly shapes the size of any network you retrieve. STRING’s Info menu has a Scores page that defines the scoring.
How to access STRING
STRING’s top-level navigation has Search, Download, My Data, Info (Scores, Use scenarios, FAQs) and Access (APIs, Usage). Which route to use depends on how much data you need.
| Route | What the Access page says |
|---|---|
| Web interface (Search) | Interactive lookup and network viewing on string-db.org. |
| REST API | Readable URLs for identifier mapping and network retrieval; documented in the developer help. Intended for limited or occasional access. |
| Download page | The route for large datasets. |
| Cytoscape app | Uses a backend API that is undocumented; STRING says to use it only for the Cytoscape app. |
STRING discourages scraping its HTML pages; use the API, the downloads or permalinks instead. For personal data sets, look under My Data in the top navigation.
Licence and reuse
All data and download files in STRING are freely available under the Creative Commons Attribution 4.0 (CC BY 4.0) licence, so reuse requires attribution.
How to cite STRING
The current STRING reference is: Szklarczyk D, et al. The STRING database in 2025: protein networks with directionality of regulation. Nucleic Acids Research. 2025;53(D1):D730-D737. doi:10.1093/nar/gkae1113. Also report the STRING version you used, because counts and scores change between releases.
Strengths and limits
- Breadth. Coverage spans more than 12,000 organisms.
- Mixed evidence. Because interactions include computational prediction and knowledge transfer between organisms, an edge is not necessarily an experimentally demonstrated interaction. Read the score and the network type (functional, physical, regulatory) before drawing conclusions.
- Volume versus confidence. Most stored interactions are low confidence (see the table above); filter by score.
- Access limits. The REST API is meant for limited use; bulk work belongs on the Download page.
Alternatives and companions
| Resource | How it differs |
|---|---|
| BioGRID | Archives curated physical and genetic interactions from primary literature, grounded in manual curation; STRING instead blends curated, aggregated and predicted associations. |
| NetworkX | A Python library for analysing network data you have exported, rather than an interaction database. |
| Biopython | A Python toolkit for scripting biological data workflows, such as processing downloaded files. |
For background, see our guides to proteomics and systems biology, the fields where interaction networks are most used.
Frequently asked questions
Is the STRING database free?
Yes. STRING states that all its data and download files are freely available under CC BY 4.0 (read November 2026).
How do I cite STRING?
Cite Szklarczyk et al., Nucleic Acids Research 2025;53(D1):D730-D737, doi:10.1093/nar/gkae1113, and name the version used.
What does the STRING confidence score mean?
It rates confidence in an interaction. As of v12.5, 953,496,894 interactions score at least 0.700 and 299,233,302 score at least 0.900, out of about 28.5 billion in total. The definition of the score is on STRING’s Scores page.
Does STRING only list physical binding?
No. It covers physical and functional associations, and version 12.5 lets you view functional, physical and regulatory networks separately.
How do I get STRING data in bulk?
Use the Download page. The API is for limited or occasional access, and scraping the site’s HTML is discouraged.
Sources
- STRING, About: string-db.org/cgi/about (read 2026-11-07)
- STRING, Access: string-db.org/cgi/access (read 2026-11-07)
- Szklarczyk D, et al. The STRING database in 2025. Nucleic Acids Res. 2025;53(D1):D730-D737. doi:10.1093/nar/gkae1113
