Reactome: A Database of Molecular Reactions and Pathways
Reactome is a free, open database of expert-reviewed human biological reactions and pathways. It records which reactions proteins, RNA, small molecules, and complexes participate in, including their cellular locations, and groups related reactions into larger pathways.
Reactome itself is not an analysis algorithm. Biological knowledge is stored in the Reactome database, while analysis tools such as the Reactome Pathway Browser and g:Profiler search that knowledge and compare it with gene lists. REAC in a g:Profiler result means a pathway category taken from Reactome.
Recording one reaction at a time
Section titled “Recording one reaction at a time”Imagine glucose moving from outside a cell into the cytoplasm. Even though both molecules are glucose, its state outside the cell must be distinguished from its state in the cytoplasm. Reactome can represent one step in which a transport protein takes extracellular glucose as input and produces cytoplasmic glucose as output as a reaction.
| Reaction element | What it means in this example |
|---|---|
| input | Molecule before the reaction, such as extracellular glucose |
| output | Molecule after the reaction, such as cytoplasmic glucose |
| catalyst | Molecule that facilitates the reaction, such as a transport protein |
| regulator | Regulatory molecule that promotes or inhibits the reaction |
| compartment | Cellular location of the molecules before and after the reaction |
In Reactome, a reaction does not mean only an enzyme reaction that changes a chemical. State changes such as molecular association and dissociation, activation, intracellular movement, and degradation are also recorded as reactions.
The relationship among entity, reaction, and pathway
Section titled “The relationship among entity, reaction, and pathway”From smallest to largest, the main Reactome units are:
| Unit | What it represents | Familiar developer structure |
|---|---|---|
| physical entity | A participant such as a protein, RNA, small molecule, or complex | Input and output values |
| reaction | One step in which input entities change or move into output entities | One function call |
| pathway | A group of related reactions | A workflow connecting several functions |
The same protein can be represented as different entities when its cellular location or modification state differs. For example, the same named protein before and after phosphorylation occupies different states within a reaction.
Reactions connect when the output of one becomes the input of the next. Reactome groups connected reactions into pathways such as Cell Cycle and DNA Repair. One reaction can belong to several pathways, so this structure is not a list of completely separate folders.
A pathway and a gene set are different
Section titled “A pathway and a gene set are different”A Reactome pathway is a biological map containing molecular states, locations, and connected reactions. A gene set used in a statistical analysis such as ORA is a list of gene IDs associated with one pathway.
| Information in the original Reactome pathway | Information retained in an ORA gene set |
|---|---|
| Order of reactions | List of genes included in the pathway |
| Input, output, catalyst, and regulatory relationships | Whether each gene is present in the list |
| Cellular locations | Usually not retained |
| Protein complexes and modification states | May be simplified into gene IDs |
A significant Reactome ORA result therefore means that the selected gene list contains more genes associated with that pathway than expected. It is not direct evidence that its reactions actually proceeded in that order or that the pathway was activated.
What you can do with Reactome
Section titled “What you can do with Reactome”Inputs and outputs depend on the question.
| Task | Input | Main output |
|---|---|---|
| Browse a pathway | Pathway name, gene ID, or protein ID | Pathway diagram connecting reactions and detailed records |
| Gene-list ORA | List of selected gene IDs | Table of pathways with more overlap than expected |
| Overlay expression values | Gene IDs and numeric values | Expression values displayed on a pathway diagram |
| Quantitative analysis of several datasets | Expression matrix and sample information | Pathway-level scores and comparison results |
The Analysis Service on the Reactome website can map input IDs to Reactome entities and run over-representation analysis. g:Profiler also provides Reactome gene sets as one data source, although its statistical correction and output column names may differ from Reactome’s own service.
Reading an ORA result table
Section titled “Reading an ORA result table”Setting sources=["REAC"] in g:Profiler runs ORA against Reactome pathways. Each output row is one Reactome pathway.
| Column | How to read it |
|---|---|
source | Data source. REAC means Reactome |
native | Stable identifier from the source, such as REAC:R-HSA-1640170 |
name | Human-readable pathway name, such as Cell Cycle |
term_size | Number of genes associated with this pathway in the background set |
intersection_size | Number of genes actually shared by the input list and pathway |
intersections | IDs of the genes that actually overlap |
p_value | Corrected probability of observing at least this much overlap by chance. Check the tool settings for the exact correction method |
A string such as R-HSA-1640170 is a Reactome stable identifier. HSA denotes the human species. Names are convenient for people but can change, so store the ID too when recording or retrieving a result.
A larger intersection_size does not automatically make a pathway more important. Large pathways naturally overlap more genes. Examine the p-value, which accounts for input-list size, background, and term_size, together with the actual overlapping genes.
Why several related pathways appear as separate rows
Section titled “Why several related pathways appear as separate rows”Cell Cycle, Cell Cycle, Mitotic, and Cell Cycle Checkpoints can all be significant. A broad pathway contains more specific child pathways, and they share genes.
Instead of counting these rows as three independent discoveries, read them in this order.
- Group pathways with similar names into the same biological theme.
- Check the genes they actually share in
intersections. - Choose whether the broad parent pathway or a more specific child pathway best fits the current question.
- Use the Reactome Pathway Browser to inspect the location and connections of the reactions.
How does it differ from GO:BP?
Section titled “How does it differ from GO:BP?”Both Biological Process in Gene Ontology and Reactome can find shared functions in a gene list, but they organize knowledge around different questions.
| Resource | Central question | Strength |
|---|---|---|
| GO:BP | Which biological processes involve these gene products? | Broad, hierarchical functional classification |
| Reactome | Which molecules proceed through which reactions into the next state? | Connected reactions, locations, and mechanistic flow |
The same cell-cycle genes may appear in both GO:BP and Reactome results. Rather than two independent experimental findings, these may be descriptions of the same gene set through two different knowledge systems.
Example: applying it to GSE251845
Section titled “Example: applying it to GSE251845”The GSE251845 exercise comparing colorectal tumor with normal tissue separated 3,631 Tumor-up genes from 4,625 Tumor-down genes and used the 24,116 genes actually tested as the background.
For Tumor-up, the top Reactome results included Cell Cycle (REAC:R-HSA-1640170) and Cell Cycle, Mitotic (REAC:R-HSA-69278). This means cell-cycle-related genes were concentrated in the upregulated list.
For Tumor-down, pathways associated with antibodies, complement, and B-cell receptor regulation appeared. This alone does not establish that immune pathways inside cancer cells were suppressed. Differences in the proportion of immune cells in tumor and normal tissue may also have affected the result. See the GSE251845 GO and GSEA exercise for the plots and settings.
Conclusions to avoid
Section titled “Conclusions to avoid”- Do not immediately call a significant Reactome pathway an activated pathway.
- Do not infer protein activity or reaction direction from RNA expression changes alone.
- Do not count parent and child pathways as independent discoveries.
- Do not mix upregulated and downregulated genes and then interpret a pathway’s direction of change.
- Do not draw a conclusion from a pathway name alone; inspect the actual genes in
intersections. - Do not automatically use every human gene as the background. Genes that could actually be tested in the experiment may be more appropriate.
- For non-human species, check whether the annotation was inferred from a human pathway.
Reactome is a database that organizes known information about proteins and reactions. It is not an instrument that directly measures cell composition, protein modifications, or actual reaction rates in an RNA-seq specimen.
What to record for reproducibility
Section titled “What to record for reproducibility”Reactome is continually updated, so record the following information.
- Reactome release or analysis date
- Analysis tool and version, such as Reactome Analysis Service or g:Profiler
- Target species
- Input ID type and mapping method
- Input-gene selection criteria and direction of change
- Background gene set
- Multiple-testing correction method and threshold
- Stable identifiers, pathway names, and actual overlapping genes
Official resources
Section titled “Official resources”- About Reactome: purpose and open scope of the database
- Reactome User Guide: pathways, reactions, and web tools
- Reactome Data Model: entity, reaction, and pathway structure
- Reactome Pathway Analysis: ID mapping, ORA, and result-table interpretation
- Reactome downloads: pathway and annotation data files