STRING

STRING (Search Tool for the Retrieval of Interacting Genes/Proteins) is a curated and predicted protein-protein interaction database maintained by the Swiss Institute of Bioinformatics, the University of Copenhagen, and other collaborating institutions. Unlike experimentally focused databases such as BioGRID, STRING integrates evidence from multiple channels — direct experimental interactions, computational predictions based on genomic context and phylogenetic profiling, text mining of scientific literature, and interaction data imported from other databases — to produce a consolidated network where each edge carries a confidence score rather than a binary existence claim.

The confidence scoring is both STRING's strength and its interpretive burden. An interaction with a confidence score of 0.9 is not necessarily direct or physical; it may reflect strong co-expression, shared pathway membership, or repeated co-mention in abstracts. Researchers using STRING for network medicine or systems biology must therefore distinguish between physical interaction networks and functional association networks — a distinction that STRING itself does not enforce but that determines whether a topological prediction is biologically meaningful or merely statistically consolidated.

STRING's integration strategy makes it one of the most comprehensive interaction resources available, but it also means that the database is a statistical synthesis of heterogeneous evidence types rather than a uniform experimental record. The confidence scores encode the reliability of the inference method, not the biological reality of the interaction under any particular cellular condition.

STRING is often used as a one-stop shop for protein interaction data, but its very comprehensiveness creates a dangerous illusion of biological completeness. A high-confidence STRING edge may never occur in the same cellular compartment, developmental stage, or post-translational state. The interactome research community needs to move from confidence-scored static maps to condition-specific dynamic networks — and STRING, for all its utility, is designed for the former, not the latter.