PSI-MI XML

PSI-MI XML (Molecular Interaction Format, MIF) is an XML-based data-exchange format for molecular interaction data, developed by the Molecular Interaction workgroup of the HUPO Proteomics Standards Initiative (HUPO-PSI). It is explicitly a data-interchange format rather than a database schema, designed to let heterogeneous interaction databases and tools compare, exchange, and verify records. The original format emerged from a consortium of academic groups and industry partners together with early interaction databases such as BIND, DIP, IntAct, MINT, and MIPS; version 1.0 dates to 2004, the widely deployed level 2.5 to 2007, and level 3.0 to 2018.

Structurally, a document is a list of entries, each grouping interactors (the participating molecules, typically proteins with cross-references to UniProt and other databases), interactions, experiments (detection method, host, publication), participants (a molecule's role within an interaction), and features (binding sites, mutations, PTMs). A distinctive design choice is heavy reliance on controlled vocabularies drawn from the PSI-MI ontology, maintained in OBO format and browsable via the EBI Ontology Lookup Service; terms cover interaction detection methods, biological roles, feature types, and more, used in either open or closed (restricted) mode. The related tab-delimited MITAB representation (format:psimi-tab-2.5, i.e. MITAB 2.5/2.6/2.7) offers a flat, quickly parsed alternative for the same conceptual model when the full XML expressiveness is not required.

Level 3.0 extends the format well beyond single-experiment description to accommodate use cases 2.5 could not: abstracted interactions assembled from multiple publications, cooperative and allosteric effects (via abstractInteraction and cooperativeEffectList), dynamic interactions whose profile varies with experimental conditions (variableParameterList), candidate molecule sets, systematic stoichiometry ranges, and richer feature handling including negative coordinates, explicit mutation sequences, and kinetic or affinity parameters tied to specific sequence changes. This makes it suitable for curated multiprotein complexes with subunit topology, as served by the Complex Portal.

In the graph-data and conversion landscape, PSI-MI XML models interactomes as attributed, provenance-rich interaction networks, contrasting with generic graph formats like format:graphml or biology-oriented pathway formats such as format:biopax-l3, format:sbml-l3v2, and format:sbgn-ml-0.3. The reference tooling is the tool:psi-jami Java library, with the tool:psicquic web service exposing IMEx Consortium data (IntAct, MINT, DIP, InnateDB, MatrixDB, and others); tool:cytoscape and tool:paxtools consume interaction data for visualization and analysis. Strengths are its ontology-grounded precision, detailed experimental provenance, and IMEx-wide adoption; the main costs are verbosity, schema complexity, and a steep learning curve that often drives casual users toward MITAB.