The Reactome pathways database¶

Reactome is a manually curated, peer-reviewed database of biological reactions. In PaintOmics it is an optional second source of pathways: it has its own explorer tab in Step 3, its own classification, its own pathway network, and one extra column in the enrichment table that no other database produces.

The application describes it on the Step 2 mapping screen as an open-source, open access, manually curated and peer-reviewed pathway database, containing information of around 20 organisms — see reactome.org. Reactome curates human and infers the other species from it by orthology, so a non-human Reactome pathway is an inference, not an independent curation.

Which organisms have it¶

Reactome is offered only where the server has installed Reactome pathways and the organism has Reactome identifier mappings. In the shipped configuration that is fifteen organisms:


Mammals	human (`hsa`), mouse (`mmu`), rat (`rno`), cow (`bta`), pig (`ssc`), dog (`cfa`)
Other vertebrates	chicken (`gga`), Xenopus tropicalis (`xtr`), zebrafish (`dre`)
Invertebrates	Drosophila melanogaster (`dme`), Caenorhabditis elegans (`cel`)
Fungi and protists	Saccharomyces cerevisiae (`sce`), Schizosaccharomyces pombe (`spo`), Dictyostelium discoideum (`ddi`), Plasmodium falciparum (`pfa`)

Whether your server has installed it is a separate question, and the Step 1 checkbox is the answer: Reactome is tickable when this deployment holds Reactome data for the organism you chose, and disabled and labelled not installed when it does not.

On your own server, Reactome must be asked for at download time (--reactome=1) and costs about 856 MB of shared data on top of the KEGG download. Reactome does not cover every KEGG organism; a species it does not cover is detected during the download, which logs a warning naming the species and carries on with KEGG data only, so the organism installs without Reactome. See Installing PaintOmics.

What a Reactome pathway contains¶

Reactome describes reactions between proteins, small molecules and complexes. PaintOmics decomposes the complexes: a node that has components is walked down to its leaf members, so the proteins inside it can each be matched and painted rather than the whole complex being skipped.

Small molecules are resolved through ChEBI to KEGG compound ids, so a metabolomics omic maps onto Reactome pathways as well as KEGG ones. Where no ChEBI-to-KEGG mapping exists, the compound is kept under its Reactome display name.

Pathway identifiers are Reactome stable ids — R-HSA- for human, R-MMU- for mouse, and so on.

Reactome pathways are painted, like KEGG's¶

PaintOmics downloads each Reactome pathway's diagram as a PNG, together with a thumbnail, and paints your feature boxes over that image using the node coordinates from Reactome's own layout data. A Reactome pathway therefore opens in exactly the same pathway view as a KEGG map, with the same boxes, the same per-condition cells and the same colour scale. The diagram is presentation only: a pathway whose image fails to download does not fail the species install.

Identifiers¶

Reactome pathways are keyed on Reactome gene identifiers (reactome_gene_id), and PaintOmics translates whatever you uploaded into that space. You do not have to supply Reactome ids — identifier translation is what the mapping step in Step 2 does — but how much of your data reaches Reactome depends on which identifier tables your organism was built with, which is why the Step 2 matrix reports the KEGG and Reactome counts side by side.

Classification¶

A Reactome pathway's classification comes from Reactome's own event hierarchy, in two levels: the top-level ancestor of the pathway and the level below it — for example Signal Transduction then Signalling by Receptor Tyrosine Kinases. Step 3's category pie and Filter by category tree are built from those two levels, exactly as they are for KEGG. A pathway for which the downloaded hierarchy holds no entry is filed as its own top level.

The Reactome class p-value column¶

When a job includes Reactome, the enrichment table gains one extra column, Reactome Class pValue.

PaintOmics groups every Reactome pathway by its top-level class, pools the genes of all the pathways in that class into a single feature set, and runs the same enrichment test on the pooled set that it runs on an individual pathway. The column shows that class's combined p-value, computed with whichever combining method the Show combined p-values control has selected, and the value is repeated on every pathway row belonging to that class. It answers a different question from the pathway's own p-value: not has this pathway moved, but has this whole branch of Reactome moved.

Three things to know before you rely on it:

It is a class value, not a pathway value. Every row in the same class carries the same number, so sorting the table by this column groups pathways rather than ranking them.
Pooling makes large classes large. A broad class such as Metabolism pools thousands of genes, and a class-level test on that set is a much blunter instrument than the test on any one pathway inside it.
The column appears whenever Reactome is one of the job's databases, so KEGG and OmniPath rows in the same table show a dash. That is not a missing value; those pathways have no Reactome class.

The class values are computed during Step 2 and stored with the job, so they are still there when you reopen the job from its URL later.

Step 3 gives Reactome its own tab¶

With more than one database in the job, Step 3 draws a tab per database. The Reactome tab has its own classification chart, its own pathway-to-pathway network and its own filters; the enrichment table below lists every database at once and can be filtered by database from its Databases to view tick-boxes. The summary band above the tabs reports how many pathways each database contributed and how many of them were significant — in the STATegra mouse example, 523 Reactome pathways of which 33 were significant, beside 364 KEGG pathways of which 71 were.

For how the four databases differ, see the comparison table on the KEGG page.