Heatmaps¶
The Global heatmap panel draws every feature of the open pathway as a heatmap, one block per omic, so you can read the whole pathway as a matrix instead of hunting for painted boxes on the map. Open it with Show Heatmap in the pathway view's toolbar; see the pathway view for the toolbar and for the diagram itself, and Feature details for the charts behind a single box.
The panel takes the same slot as the Feature set overview, so opening one closes the other. Its header carries hide, configure (the cogs), expand and shrink controls, and its left edge can be dragged to widen it.
Choosing what to draw¶
The configurator is open when the panel first appears, and nothing is drawn until you press Apply. The cogs icon reopens it later; it slides shut again each time a drawing finishes.
Under Choose the omics to draw there is one entry per omic that has at least one matched feature in this pathway — an omic that matched nothing here is not offered. Each entry has:
| Control | What it does |
|---|---|
| The checkbox | Whether this omic gets a heatmap at all. All are ticked to begin with. |
| All features (Genes or compounds) | Draw a row for every feature of this pathway that this omic measured. |
| Only relevant features | Draw only the rows that are relevant, or that carry a relevant association, for this omic. This is the initial choice. |
| The numbered handle | Drag the entry up or down to change the order the heatmaps are drawn in. The numbers renumber as you drop. |
The order matters for more than layout: with Force order for features switched on, the omic at the top of the list is the one every other heatmap follows.
An omic with nothing left to draw — Only relevant features chosen and nothing relevant in this pathway — shows "No data" in place of its heatmap.
Advanced options¶
This can take up to ten seconds
The panel warns you here in its own words: depending on the settings chosen, generating the heatmaps can take up to ten seconds. A pathway with hundreds of matched features, several omics and clustering switched on is the slow case; Only relevant features is the fast one.
Force order for features¶
Off by default. Each heatmap is then clustered and ordered on its own, and the row at a given height in one omic has nothing to do with the row at that height in the next.
Switched on, only the first omic in the list is ordered. Every other heatmap is redrawn in that omic's row order, so a row sits at the same height in all of them and a profile can be followed straight across. Two consequences follow from that, and both are visible on screen:
- A feature the reference omic has but a secondary omic does not still gets a row there, drawn empty and labelled NO DATA, to keep the rows aligned.
- A secondary omic will draw a row that is not relevant for it, if the reference omic contributed that row. The shared order takes precedence over that omic's Only relevant features setting.
Show per-condition significance stars¶
On by default. A star on a cell means the feature is relevant for that condition — that your relevant-features file listed it there. The star's ink is chosen against the cell's own colour so it stays legible on the pale middle of the scale.
Two limits are worth knowing. Stars are only drawn when the omic has more than one column, and a relevant-features file that is a single column of identifiers means "relevant overall": it produces no per-cell stars anywhere, only the star in the row label. Per-condition stars need a per-condition relevance file, one column per value column. Turning the checkbox off is useful when a heatmap is dense enough that the stars fight with the colours.
Clusterize data¶
On by default, with Hierarchical clustering selected. Clustering is skipped, whatever the setting, for a heatmap with fewer than three rows.
Hierarchical clustering groups the rows and draws a dendrogram to the right of the heatmap. The tree is computed on the rows themselves: each row is the vector of values you can see in it — one number per column, in the replicate/sample mode currently selected — compared by euclidean distance and merged by complete linkage. The leaves are reordered so that similar profiles end up adjacent, which is what makes blocks of common behaviour visible.
Read the dendrogram for its grouping, not for its distances. Branches are drawn one step per level of the tree, so the horizontal position of a join records its depth in the tree, not how far apart the two groups were.
K-means clustering partitions the rows into k groups instead, and marks them with a numbered coloured bar down the right-hand side, one per cluster, rather than a dendrogram. A box appears per omic so you can set its own k. The bar colours are generated per drawing and mean nothing beyond "these rows are in the same cluster"; they are not stable between redraws.
Reading a heatmap¶
Rows. The label is the feature's symbol with its KEGG identifier on the first line and the identifier you uploaded on the second. A long label is shortened from the end that carries least: the symbol keeps its beginning, the uploaded identifier keeps its ending, which is where identifiers that share a prefix differ. Hover the label for both in full. A red star marks a relevant feature and a gold star a relevant association.
For an omic uploaded with associations — transcription factors, miRNAs,
methylation, anything using the target:::regulator input form — the two lines
are swapped: the row is identified by the regulator, with the target
underneath. It is the inverse of an ordinary omic, where the row is the gene.
See Regulatory omics.
Hovering a row outlines the same feature's row in every other omic's heatmap, which is the quickest way to ask whether a gene that moves in the transcriptome moves in the proteome too. The linkage follows the target, so a regulator row lights up with its target's gene-expression row.
Columns are labelled with the condition names from that omic's own file header, rotated and length-capped, with the full name in the cell tooltip. Where the header cannot be matched to the columns being drawn, the labels fall back to "Condition 1", "Condition 2" and so on rather than risk naming a column wrongly.
Colours come from the same settings as the painted diagram: each omic is coloured against its own reference range, chosen in the Settings panel, and a legend under each omic heading shows the two ends of that range, a tick at zero where the range crosses it, and a caption naming the reference in words — "10th-90th percentile" by default. Changing a colour setting and pressing Apply in Settings repaints this panel along with the diagram. On the default Blue-Grey-Red scale a value past either end of the range is painted darker rather than clipped, so an extreme value still reads as extreme.
What this panel does not do¶
There is no download here: the heatmaps cannot be exported as an image or as a table, and the only export in the pathway view is the diagram's own Download button, which writes a PNG of the map. If you need the numbers themselves, they are in the files you uploaded, and the matched feature lists are in the Pathway information panel.