Skip to content

Distribution charts

The Distribution section of the Spatial omics panel charts whatever the map is coloured by, so the chart and the map always agree. When observations are selected, the charts show or highlight the selection.

The header names what is charted ("gene Ttr", or the column name), with a "n selected" badge while there is a selection. Each chart type answers a different question. The beside the tabs explains the one on screen, including what it must not be read for.

Chart Offered when the map is coloured by Shows
Counts a group (categorical column) How many observations in each category
Histogram a gene or measurement How the value is distributed
Violin a gene or measurement The value's distribution per group, as a smoothed density
Box a gene or measurement Median, quartiles and range per group
Heatmap always Mean expression of picked genes per group
Embedding the dataset publishes an embedding (UMAP, PCA, t-SNE) The cells in the embedding space

The chart switches to Counts when you colour by a group, and to Histogram when you colour by a value (unless Heatmap or Embedding is showing). With no colouring, the panel asks you to colour the map by a column or a gene first.

Grouping (Violin, Box, Heatmap): pick a categorical column to split by. "No grouping" draws one trace.

Counts

Counts chart

Horizontal bars, largest at the top. The 25 largest categories are shown; the rest are folded into "other (N categories)" in grey. With a selection, the full counts (All) are drawn faint, with the Selected counts over them. A category is a label, not a magnitude, so counts are the only distribution it has.

Histogram

Histogram with a selection overlaid

The number of bins is the square root of the number of observations, between 10 and 100. The x axis is the value, or log1p(value) when Log scale is on. With a selection, the Selected histogram is overlaid on the full one (All), not replacing it.

Violin and Box

Violin split by cluster

One violin or box per group, in the map's colours (empty groups are dropped). Violins show an inner box and the mean. With a selection, only the selected observations are charted ("Showing the selection only."). Each trace is thinned to 20,000 points for speed.

  • Violin shows the shape: two groups with the same median can have one peak or two.
  • Box shows median, quartiles and range, but hides whether a group is bimodal.

Heatmap

Heatmap of picked genes per cluster

Mean expression of each picked gene (rows) within each group (columns).

Control Default What it does
Pick genes for the rows the gene the map is coloured by One row per gene. No limit on the number of genes.
Grouping the first categorical column The column the heatmap's columns come from.
z-score on Scales each gene across the groups, so a colour means "above or below this gene's own average". Off: raw means, so one highly expressed gene can saturate the scale.
  • Up to 40 group columns are shown (the strongest); the note says when some were left out.
  • With a selection of up to 200 cells, each selected cell gets its own column.
  • A group needs at least 3 measured cells to get a mean.
  • Colours: blue–white–red around 0 when z-scored; white to dark blue for raw means. Missing values are gaps.

All genes

When the dataset and the server support it, a Picked genes / All genes switch appears. All genes draws every gene against every cell as a zoomable image:

Control Choices Default
View Cells (a column per cell, sorted by group) / Groups (each gene's mean per group) Cells
Scale (Cells) Expression / Per gene (0–1) Expression
Scale (Groups) z-score per gene / Mean expression / Per gene (0–1) z-score per gene
Detected genes only Leave out genes never detected on
Find a gene Exact, then prefix, then substring match —
Reset zoom — —

Scroll to zoom (Shift: genes only, Alt: cells only), drag to pan, double-click to reset. Click a gene to colour the map by it. Zoomed out, a pixel is the mean of a block of cells and genes; zoom in for single cells and gene names. The first time for a grouping takes one pass over the expression matrix.

Embedding

Embedding (UMAP) with a lasso selection

The cells in an embedding space (UMAP, PCA or t-SNE), coloured like the map when it is coloured by a group. Choose the embedding in the dropdown when the dataset has several. 3D embeddings can be rotated.

  • Lasso or box-select points to select them on the map: both views read the same selection. An empty lasso clears it. Unselected points are dimmed.
  • Only PCA labels its axes with a percentage, because only its axes explain a measurable share of the variance. UMAP and t-SNE axes are arbitrary: do not read the distance between distant clusters, or the direction of an axis.
  • A legend is drawn for up to 10 categories.

Computing t-SNE in the browser

For a dataset with a PCA, no t-SNE and at most 5,000 observations, the dropdown offers t-SNE (compute) and t-SNE 3D (compute). Click Compute (the button shows the estimated time) to run it in the background on your GPU; the panel stays usable and Cancel stops it. The result is not saved: a reload computes it again. Larger datasets say t-SNE is not offered, because it would take too long in the browser.

Detach

Detach moves the chart into its own window, so it can sit beside the map. The panel then says "Shown in its own window." The window can be moved and resized; switching chart type keeps it where it is. Put back , or closing the window, returns the chart to the panel.

A chart detached into its own window