Exploration of gene expression variation across cell types is easily achieved for the typical clustered embedding plots. Visual inspection along other variables of interest (such as drug treatment, time point, cell cycle, etc.) in a clustered embedding plot is less straightforward because the cells from different states in the variable of interest are grouped by cell type causing datapoints from different states to overlap.

Use plot filtering to drill down into the data by isolating subsets of cells based on available experimental metadata. 

For example, how does TNC expression vary over the course of lactation (milk_stage)?

A single plot of TNC gene expression can't answer the question because cells at different stages will plot to the same areas when the data is clustered to determine cell type. By clicking “Filter plotted cells”, subsets of cells can be selectively displayed. To visualize TNC expression for cells early in the course of lactation, select the filter "early" under milk_stage. (Visit Filter plotted cells documentation for more details)

In contrast, there is relatively little TNC expression for cells in the late stages of lactation (late 2, late 3, late 4).

Examining a slightly earlier stage of lactation (late 1), an intermediate amount of TNC expression is seen.

Subset on a separate experimental variable to explore whether hormonal birth control had any effect on TNC expression, by combining filters for milk_stage: late1 and Hormonal birthcontrol: yes

Compare the above visualization with the results from filter combination milk_stage:late1 and Hormonal birthcontrol: no (below)

Plot filtering is a new feature that SCP plans to extend. Future work includes applying plot filtering to:

• numeric annotations
• distribution plots
• spatial studies
• dot plots

As a study owner, if plot filtering is not available for your study, please email scp-support@broadinstitute.zendesk.com for help. 

Have questions, comments or suggestions about plot filtering? Contact us at scp-support@broadinstitute.zendesk.com to share your feedback.