Scatter plots

Scatter plots for cell ordination

Often a first question when performing single-cell analysis is which cells are similar to each other. This similarity can be for many reasons including similar cell type, state, or other biological and technical reasons. Many different methodological techniques can result in hard and soft clustering of cells based on underlying high-dimensional measurements. In the portal, interactive scatter plots are available to explore cell ordination. Both 2 and 3-dimensional plots are available in the portal. Also available are correlation plots assessing coordinated expression between pairs of genes and spatial plots to visualize the physical scatter of gene expression.

2D plot

Here is an example of a 2D scatter plot of cells from Brain MD70. Although a picture is shown here, the live interactive plot can be found here.

3D plot

Here is an example of a 3D scatter plot of cells showing a trajectory (diffusion components) analysis derived from data from a Small Intestinal Epithelium study. Although a picture is shown here, the live interactive plot can be found here (please select "Regional Diversity of IECs (Figure 2c)(DC)" in view options found on the upper right).

Correlation plot

In some cases, it is useful to know whether a group of genes tends to be co-expressed -- that is, the genes are highly and lowly expressed in the same cells. Identifying a group of co-expressed genes might indicate that those genes are co-regulated, acting as part of a common mechanism. One way to identify co-regulated genes is to calculate the correlations between all pairs of genes, and then clustering the results to find distinct groups of genes with a common expression pattern.

In some cases, study owners upload a pre-computed heatmap to show these clustered correlations between genes. If not, you can still check the correlation between two genes by looking at a correlation scatter plot. This can help you understand whether two genes of interest are likely to be part of a larger set of co-regulated genes. In addition, you can get a sense for whether the correlation might change depending on which cells you examine; for example, two genes might be co-regulated in one cell type but not in another.

Here is an example of a correlation scatter plot comparing the genes Lyz1 vs. Mptx2 from Study: Small intestinal epithelium. Although a picture is shown here, the live interactive plot can be found here.

Note that the Spearman's correlation value displayed above the scatter plot may be inflated by the data's normalization procedure, and therefore may not accurately reflect the data. The normalization may also introduce artifacts into the plots themselves -- in particular, data arranged in perfectly-straight lines (as shown in the example plot above), are likely the result of an artifact and should be interpreted with caution.

Spatial plot

Here is an example from a Visium demo study with a cluster plot on the left and a spatial plot to the right. Although a picture is shown here, the live interactive plot can be found here.