STRING

STRING
Cvmering · CC BY-SA 4.0 · source
Name	STRING
Developer	European Molecular Biology Laboratory EMBL Center for Genomic Regulation Swiss Institute of Bioinformatics Max Planck Society
Initial release	2000
Latest release	2025
Programming language	C++ / JavaScript / Python
Operating system	Cross-platform
License	Academic / commercial licensing

STRING

The STRING database is a widely used bioinformatics resource that aggregates known and predicted protein–protein interactions for numerous organisms. It provides a consolidated interface to explore physical associations and functional links among proteins drawn from experimental data, curated databases, computational prediction methods, and text mining. Researchers in molecular biology, systems biology, structural biology, and biotechnology commonly use it alongside resources such as UniProt, PDB, Gene Ontology, Ensembl, and KEGG.

Introduction

STRING is a centralized interaction network resource integrating interactions from diverse sources including experimental repositories, curated pathway collections, and computational predictions from orthology-based transfer and text mining. The platform links protein identifiers from projects like UniProt, Ensembl, RefSeq, HGNC, and NCBI to interaction evidence drawn from databases such as BioGRID, IntAct, Reactome, MINT, and DIP. STRING’s web portal, programmatic APIs, and downloadable network files are used in conjunction with analysis tools such as Cytoscape, Gephi, R Project packages, and Python (programming language) libraries for network visualization and enrichment analysis.

History and Development

Development of STRING began with collaborative efforts at computational biology centers including European Molecular Biology Laboratory groups and partners like Swiss Institute of Bioinformatics. Early versions consolidated curated interactions from resources such as BIND and DIP before expanding to systematic text-mining pipelines and orthology-based transfer from model organisms like Saccharomyces cerevisiae, Drosophila melanogaster, Caenorhabditis elegans, Mus musculus, and Homo sapiens. Over successive releases the project incorporated data from large-scale projects and consortia including Human Genome Project outputs, ENCODE, GTEx, and pathway efforts like KEGG and Reactome. Funding and institutional support have come from agencies and entities such as the European Commission, Human Frontier Science Program, and national research councils linked to institutions like the Max Planck Society.

Data Sources and Integration

STRING integrates multiple evidence channels: curated pathway databases (Reactome, KEGG), experimental repositories (BioGRID, IntAct), genomic context analyses drawn from comparative genomics projects tied to Ensembl and RefSeq, and automated text mining of literature indexed in PubMed and publisher archives. Orthology-based transfer of interactions uses resources and standards from OrthoDB and homology mapping informed by sequences in UniProt and structural inferences from PDB. Metadata harmonization references identifier resources such as HGNC for human gene symbols and taxonomy backdrops from NCBI Taxonomy. Integration workflows parallel pipelines used in projects like STRING-DB predecessors and align with data-exchange initiatives exemplified by BioPAX and PSI-MI.

Scoring and Confidence Metrics

Each reported interaction is accompanied by quantitative confidence scores reflecting evidence strength from channels including experimental data (BioGRID), curated pathways (Reactome), gene neighborhood and gene fusion signals derived from comparative projects involving Ensembl genomes, and co-expression evidence from transcriptome collections such as GTEx and ArrayExpress. Scores are combined into a probabilistic framework similar to approaches used by integrative projects like ConsensusPathDB and are presented as normalized confidence values enabling thresholding for downstream analysis with tools like Cytoscape and statistical packages from the R Project. Benchmarking and calibration of scores have been performed against gold-standard sets compiled from pathway resources like KEGG and curated complexes described in CORUM.

User Interface and Access Methods

Users access STRING via a web interface developed with modern web stacks and interactive visualization components compatible with Cytoscape and export formats used by GraphML and SIF. Programmatic access is provided through RESTful APIs and downloadable datasets for bulk analysis in environments such as R Project Bioconductor packages and Python (programming language) scripts. Integration with workflow platforms like Galaxy (software) and interoperability with network analysis ecosystems exemplified by Cytoscape apps facilitates incorporation into pipelines used by laboratories at institutions such as European Bioinformatics Institute and university core facilities.

Applications and Use Cases

STRING is applied in functional annotation of novel proteins, hypothesis generation for experimental design in labs at institutions like Harvard Medical School and Stanford University School of Medicine, network-based interpretation of omics studies from consortia such as The Cancer Genome Atlas and ENCODE, and drug-target network exploration in collaborations with pharmaceutical groups including Novartis and Roche. It supports studies of conserved interaction modules across taxa including Saccharomyces cerevisiae, Escherichia coli, Arabidopsis thaliana, and Homo sapiens, and is used in pathway enrichment workflows that link to resources like Gene Ontology and KEGG.

Limitations and Criticisms

Critiques of STRING center on potential propagation of annotation transfer errors via orthology-based inference between species such as Mus musculus and Homo sapiens, biases introduced by literature-derived text mining that favor well-studied proteins tied to institutions like NIH-funded research programs, and variable quality among source databases including legacy entries from BIND. Users also caution about over-reliance on aggregate confidence scores for mechanistic interpretation without experimental validation typical of studies published in venues like Nature and Science. Methodological transparency, versioning, and provenance tracking remain important topics echoed in discussions at forums such as ISMB and in guidelines from organizations like FAIR initiatives.

Category:Bioinformatics databases