Interactions · Database
STRING
Database of known and predicted protein-protein interactions and functional association networks across organisms.
What STRING is
STRING (Search Tool for the Retrieval of Interacting Genes/Proteins) is a biological database and web resource containing known and predicted protein-protein interactions. Developed by an academic consortium including CPR, EMBL, KU, SIB, TUD, and UZH, it integrates experimental records, computational predictions, and literature mining into unified, species-specific association networks.
The database addresses the challenge of mapping complex cellular pathways and functional partnerships across diverse organisms. By assigning confidence scores to interactions and enabling cross-species propagation via orthology, it assists researchers in interpreting genomics data and understanding system-level cellular processes.
The platform encompasses functional association networks, physical subnetworks, and regulatory subnetworks spanning tens of thousands of organisms, supported by tools for functional enrichment analysis and whole-proteome annotation.
How it is organised
STRING imports experimentally validated interactions and curated pathway annotations from sources including MINT, HPRD, BIND, DIP, BioGRID, KEGG, Reactome, IntAct, EcoCyc, GO, and the NCI-Nature Pathway Interaction Database. Computational predictions are generated through multiple channels: text mining of scientific literature (such as PubMed, OMIM, SGD, and FlyBase) for gene co-occurrences, genomic context evaluations (conserved gene neighbourhood, gene fusion and fission events, and phylogenetic co-occurrence profiles), co-expression patterns, and orthology-based interaction transfers from model organisms.
All imported and predicted associations are benchmarked against KEGG functional pathway references and weighted to produce a combined confidence score. Data can be viewed in either protein mode or COG mode (Clusters of Orthologous Groups). Users can interrogate associations through a web interface, programmatic API calls, bulk downloads, or a Cytoscape plug-in, setting confidence cut-offs (low, medium, high, or highest) and network types.
Key facts
- Latest version
- 12.5
- Number of organisms
- 12,535
- Number of proteins
- 59.3 million
- Number of interactions
- >20 billion
- Confidence score tiers
- Highest (0.900), High (0.700), Medium (0.400), Low (0.150)
- Supported network types
- Full STRING network, physical subnetwork, regulatory subnetwork
- Developer consortium
- CPR, EMBL, KU, SIB, TUD, UZH
- Access methods
- Web interface, API, bulk download, Cytoscape plug-in
What you can use it for
- Protein interaction network exploration
- Query single proteins, protein pairs, or lists of proteins by name, identifier, or amino acid sequence to generate interactive interaction maps with adjustable confidence thresholds.
- Functional enrichment analysis
- Submit ranked experimental protein lists (such as fold-change or abundance metrics) to test for enrichment across Gene Ontology categories, KEGG pathways, Pfam domains, and disease annotations without pre-filtering.
- Custom proteome annotation
- Upload a full newly sequenced species proteome to automatically infer functional networks and predict functional annotations such as Gene Ontology terms and KEGG pathways.
- Pathway and disease gene-set visualization
- Search by pathway name, disease, or scientific publication to visualize interconnected protein sets and assess interconnectivity or shared molecular functions.
Query it here
The research agent queries STRING directly — no account, no API key — and every answer links back to the record it came from. Ask in plain language and it writes the query.
Strengths and limits
Good at
- Integrates both direct physical bindings and indirect functional associations across thousands of species.
- Provides calibrated confidence scores benchmarked against standardised KEGG pathway data.
- Enables transfer of interaction knowledge across species via orthology inference.
- Offers flexible input methods including protein identifiers, amino acid sequences, COG families, and ranked experimental values.
Watch out for
- Many reported associations represent predicted or text-mined linkages rather than experimentally confirmed physical interactions.
- Annotation depth and interaction coverage vary by organism, requiring orthology transfers from well-studied model systems.
Questions
Can I search STRING without writing code?
Yes. The research agent queries STRING directly, with no account or API key, and cites the record behind every answer. Ask in plain language — for a gene, a protein, a compound or a study — and it builds the query for you.
Can I reuse data from STRING?
CC BY 4.0. Corollary filters and normalizes associations. Scores are confidence, not affinity. Terms: https://string-db.org/cgi/access?footer_active_subpage=licensing
What is the difference between functional associations and physical subnetworks in STRING?
The full functional network includes both direct physical interactions and indirect functional associations (such as shared pathways, co-expression, or genomic context links), whereas the physical subnetwork restricts edges strictly to binding or physical complex partnerships.
How are confidence scores defined?
Interactions receive a score based on supporting evidence benchmarked against KEGG reference data, grouped into standard cut-offs: low confidence (0.150), medium confidence (0.400), high confidence (0.700), and highest confidence (0.900).
Can I analyse unannotated sequences or newly sequenced species?
Yes. STRING allows queries using raw amino acid sequences and provides a tool to upload complete custom proteomes to generate predicted interaction networks and functional annotations.
Written from STRING's own website and reference material. Statements citing a figure that does not appear there are removed before publication. Licence and running details come from our registry.