The OmniPath interaction network¶
OmniPath is the fourth pathway source PaintOmics can run, and the only one that ships no diagram. It integrates over 100 resources into a prior-knowledge network of signed, directed molecular interactions — A stimulates B, C inhibits D — for human, mouse and rat. Because there is no drawn map to paint, an OmniPath pathway opens as an interactive interaction network instead.
The application describes it on the Step 2 mapping screen in those terms and links to omnipathdb.org.
Which organisms have it¶
Three, and only three: human (hsa), mouse (mmu) and rat (rno). The
omnipathdb.org web service serves those taxa and rejects every other, so no
other organism can have OmniPath installed however the server is configured.
Whether your server has installed it for one of the three is what the Step 1 checkbox says. Installing it is a separate command from the ordinary organism install — see Installing PaintOmics.
What an OmniPath "pathway" is¶
OmniPath keeps pathway membership and the interaction network in two separate datasets, and PaintOmics combines them:
- Membership comes from OmniPath's curated annotations, restricted to two resources: SIGNOR and NetPath. SIGNOR contributes focused causal modules (median around 18 genes); NetPath contributes broader receptor cascades (median around 92).
- Edges come from OmniPath's interactions endpoint — the core literature-curated set plus its directed and kinase-substrate extensions. The transcriptional and miRNA layers are deliberately left out: they are a different kind of statement and would swamp the graph.
- A pathway is the subnetwork its members induce on that graph. A pathway with fewer than five members present in the network is dropped, because the subnetwork would be too small to read and too small to enrich against.
Mouse and rat need one extra step. OmniPath's annotations are human, so membership is carried across using the ortholog pairs OmniPath itself used, recovered by aligning the mouse or rat interaction table against the human one on their shared source literature — not guessed from gene-symbol casing.
In the STATegra mouse example the job found 120 OmniPath pathways, 10 of them significant, beside 364 from KEGG and 523 from Reactome.
What you see when you open one¶
Click Paint on an OmniPath pathway and the pathway panel draws a graph you can move, zoom and interrogate, rather than a picture.
- The nodes are your data. Each is the same painted glyph a KEGG map would place on its diagram — one box per matched gene, split into one coloured cell per condition, on the same Blue-Grey-Red scale. A gene looks identical here and on a KEGG map.
- The edges carry the sign, which is the thing OmniPath knows and a diagram does not. A green arrow is stimulation, a red bar-headed edge is inhibition, and a plain grey line with no arrowhead is an interaction with no recorded direction of effect. The key is in the toolbar.
- Layout offers three arrangements: Rings (by connectivity), which puts the pathway's hubs in the middle and cannot overlap; Cascade (follows direction), which runs the graph top-down from the nodes nothing points at, the way a signalling diagram is usually drawn; and Clusters (force-directed). Rings is the default above about 45 nodes and Clusters below it.
- Causal edges only hides the unsigned edges.
- Hover a node and everything outside its immediate neighbourhood fades, so you can read one gene's inputs and outputs out of a dense graph. Click a node and the feature detail panel opens, the same one a click on a painted KEGG box opens.
- The status line reports how many genes and interactions are drawn. A pathway is capped at 900 edges; when the cap bites the line says so — (strongest 900 of N) — and signed edges are always kept in preference to unsigned ones.
Warning
Node positions carry no biological meaning. OmniPath publishes no layout, so every position you see was computed — by the layout you chose, from the connectivity of the graph. Two nodes drawn close together are not thereby related; read the edges, not the geometry. This is also why the pathway is presented as a graph you can rearrange rather than as a fixed map.
Identifiers¶
OmniPath is keyed on UniProt accessions, with RefSeq gene symbols as the alternative when you upload names rather than accessions. You do not have to supply either: identifier translation converts what you uploaded, and where an identifier cannot reach UniProt directly it is bridged through gene-level identifiers — Entrez/NCBI gene ids, Ensembl gene ids, KEGG gene ids — which is how an Ensembl gene id in your file ends up on an OmniPath node.
OmniPath has no metabolites¶
Every OmniPath pathway is built with an empty compound list. Nothing from a metabolomics omic can match an OmniPath pathway, and no OmniPath enrichment p-value is ever computed from compounds.
One place in the interface does not show this. In the Step 2 Multiple databases used matrix, a compound-based omic shows the same count in every database column, because compounds are matched once against KEGG compound ids and the server has no per-database compound breakdown to report. The OmniPath cell on a metabolomics row is therefore not zero, but no OmniPath pathway can use it. Gene-based rows in that matrix are genuine per-database counts.
Classification¶
OmniPath publishes no pathway hierarchy — SIGNOR and NetPath are flat lists of names — so the classification you see is one PaintOmics assigns, not one carried from upstream. It has two levels.
The top level is a category matched from the pathway's name, first match winning, so a pathway named for a tumour is filed under cancer even though it is also a signalling cascade:
| Category | Filed here when the name mentions |
|---|---|
| Cancer | a tumour type or an oncogenic process |
| Infection and inflammation | a virus, an infection, inflammasomes, complement, innate immunity, interferon |
| Immune signalling | B or T cells, macrophages, interleukins, chemokines, TNF, RANKL |
| Nervous system | Alzheimer, Parkinson, axons, synapses, a neurotransmitter system |
| Cell cycle, death and autophagy | the cell cycle, apoptosis, autophagy, death receptors, differentiation |
| Metabolism | metabolism, biosynthesis, insulin, leptin, AMPK, circadian |
| Development and tissue remodelling | fibrosis, Hedgehog, Wnt, Notch, integrins, focal adhesion |
| Signal transduction | everything else — these are all receptor-to-nucleus cascades |
The second level is the curating resource, SIGNOR or NetPath.
In the Step 3 pathway network for OmniPath, two pathways are joined when they share genes, weighted by how many, and the pathways are grouped under their curating resource.
Pathway identifiers and external links¶
OmniPath pathway ids are slugs PaintOmics mints from the resource and the pathway name; they are not accessions any external site knows. The enrichment table's external-link button therefore opens the annotation query for that pathway's resource rather than a record for the pathway itself.
For how the four databases differ, see the comparison table on the KEGG page.