BioGRID

BioGRID (Biological General Repository for Interaction Datasets) is a freely available, manually curated database of physical and genetic interactions, chemical associations, and post-translational modifications drawn from the primary biomedical literature. It was established in 2003 and is maintained by the Tyers Lab, with development spread across the Hospital for Sick Children and Princeton University; funding has come from the NIH, the Canadian Institutes of Health Research, and related agencies. Content is compiled by professional curators from tens of thousands of publications and updated on a roughly monthly release cycle. Recent releases catalog on the order of three million protein and genetic interactions, over a million PTM records, and tens of thousands of chemical associations across humans, major model organisms (yeast, fly, worm, mouse, Arabidopsis), and viral pathogens such as SARS-CoV-2 and HIV.

The core data model is an interaction record linking two interactors (genes or proteins, identified by systematic and standard names plus cross-references to Entrez, UniProt, and other databases), annotated with an experimental detection method, throughput class (low- versus high-throughput), a supporting PubMed reference, and, where relevant, quantitative scores. Detection methods are mapped to the community PSI-MI controlled vocabulary, which lets BioGRID export interoperable files. Its native download format is the tab-delimited BioGRID TAB 3.0, a spreadsheet-friendly columnar layout; the project also publishes the standardized format:psimi-tab-2.5 (MITAB) flat format and format:psi-mi-xml-3.0 XML, alongside project-specific subsets and PTM site files.

Within the graph-data and format-conversion landscape, BioGRID functions as a source of curated interaction networks rather than a graph engine. Data can be retrieved through the web interface, bulk archive downloads, or a URL-parameterized REST web service that filters by gene, PubMed ID, evidence type, and organism. A sibling resource, BioGRID-ORCS, captures gene-level phenotypes and genetic interactions from genome-wide CRISPR/Cas9 screens. Downloaded interactions are commonly loaded into network tools such as tool:cytoscape, tool:gephi, tool:networkx, or tool:neo4j, or accessed programmatically via the tool:biogridpy client; the PSI-MI outputs also interoperate with standards-based tooling like tool:psi-jami and tool:psicquic.

Its strengths are curation depth, breadth of organisms and evidence types, transparent provenance to source publications, and standards-compliant exports. Limitations include the absence of a native graph query language, the aggregation of interactions from heterogeneous assays of varying reliability (so evidence codes must be filtered carefully), literature-driven coverage bias toward well-studied genes, and the fact that TAB 3.0 remains a bespoke format requiring conversion for many graph pipelines.