Reactome

Reactome distinguishes itself through expert biocuration rather than computational inference. Each pathway is authored by PhD-level biologists, peer-reviewed, and cross-referenced to primary literature. This manual process captures reaction-level detail including:

• Input and output molecules (proteins, small molecules, nucleic acids)
• Catalytic activities and stoichiometry
• Post-translational modifications
• Compartmentalization within cellular organelles
• Disease-associated variants and mutations

Reactome organizes biological knowledge in a nested hierarchy that mirrors cellular architecture. The top level consists of 27 superpathways (e.g., Signal Transduction, Metabolism, Immune System), which decompose into subpathways and ultimately into individual reaction events. Each reaction is a discrete molecular transformation with defined inputs, outputs, and catalyst entities. This hierarchical structure enables:

• Granular querying at any resolution level
• Aggregation of reactions into higher-order biological processes
• Systematic over-representation and enrichment analysis against the full hierarchy

Every physical entity in Reactome is mapped to authoritative external databases, creating a semantically interoperable knowledge graph. Key cross-references include:

• UniProt for protein sequences and functional annotations
• ChEBI for small molecules and metabolites
• Ensembl and NCBI Entrez Gene for genomic coordinates
• Gene Ontology (GO) for molecular function and biological process terms
• PubMed for literature evidence supporting each reaction This dense cross-referencing enables seamless integration with multi-omics analysis pipelines and programmatic access via RESTful APIs.

Reactome provides a built-in over-representation analysis (ORA) tool that identifies pathways significantly enriched in user-supplied gene or protein lists. The analysis employs a hypergeometric distribution test with Benjamini-Hochberg false discovery rate (FDR) correction. Key outputs include:

• Enrichment p-values and FDR-adjusted q-values
• Fold enrichment ratios
• Interactive pathway diagrams with differentially expressed entities highlighted
• Hierarchical visualization of enriched pathways as a navigable tree
• Exportable results in tabular and graphical formats for publication

Reactome curates pathways primarily for Homo sapiens but computationally projects these annotations onto orthologous proteins in 15 additional model organisms including:

• Mus musculus (mouse)
• Rattus norvegicus (rat)
• Danio rerio (zebrafish)
• Drosophila melanogaster (fruit fly)
• Saccharomyces cerevisiae (budding yeast) This orthology-based inference uses Ensembl Compara protein orthologs, enabling translational researchers to map findings from model organisms onto human pathways and vice versa.

Reactome supports computational workflows through multiple programmatic interfaces:

• Content Service REST API: Retrieve pathway diagrams, reaction details, and entity information in JSON format
• Analysis Service REST API: Submit gene lists for enrichment analysis and retrieve results programmatically
• BioPAX Level 3: Export pathways in the standard biological pathway exchange format
• SBGN-ML: Export in Systems Biology Graphical Notation markup for visualization tools
• MySQL Database Dumps: Download the complete Reactome relational database for local analysis
• Graph Database: Access the Reactome knowledge graph via Neo4j for network-based queries

A computational method that determines whether a priori defined gene sets show statistically significant, concordant differences between two biological states. Unlike Over-Representation Analysis, GSEA uses a running sum statistic that walks down a ranked gene list—increasing when a gene belongs to the Reactome pathway and decreasing otherwise—to detect non-random clustering at the extremes. The resulting Enrichment Score (ES) reflects the maximum deviation from zero, and phenotype permutation preserves gene-gene correlations when estimating the null distribution.

An integrated database resource consisting of manually curated pathway maps representing molecular interaction and reaction networks. While Reactome provides detailed mechanistic reaction steps with physical entities, KEGG PATHWAY offers broader metabolic and signaling overviews. KEGG's orthology-based annotation links genomes to pathways across species, and its BRITE functional hierarchy complements Reactome's event-based structure for multi-database enrichment strategies.

A structured, species-independent bioinformatics framework providing a controlled vocabulary across three domains: biological process, molecular function, and cellular component. Reactome pathways map to GO terms via cross-references, enabling unified enrichment analysis. GO's directed acyclic graph structure supports semantic similarity calculations that quantify functional relatedness between terms based on shared ancestor information content—a technique often applied alongside Reactome pathway clustering.

A comprehensive collection of annotated gene sets including Reactome pathways as one of its C2 curated gene set collections. MSigDB's Hallmark gene sets—50 refined biological states generated by reducing redundancy across founder sets—are frequently used alongside Reactome for orthogonal validation. The database stores collections in GMT file format, where each row contains a gene set name, description, and list of identifiers, enabling direct loading into GSEA software.

An enrichment approach that incorporates structural dependencies, interaction types, and signaling directionality from Reactome's molecular network into statistical assessment. Unlike ORA or GSEA—which treat genes as independent entities—topology-based methods weight genes by their network position (e.g., upstream regulators vs. downstream effectors) and account for pathway crosstalk where overlapping gene membership between pathways can confound independence assumptions. Tools like Pathview overlay experimental data directly onto Reactome diagrams.