alexvpickering
commented
11 months ago This sounds like a great idea Sara. I would adjust the language for the tooltip slightly:
- The bolded part is a must (can't say they are not significant -- what the statistical test tells us -- prior to running the statistical tests)
- it's not just all zeros that will be excluded (see Filtering to remove low counts section in linked article from
?edgeR::filterByExpr). Your phrasing is fine too as you used that as an example
For between-group comparisons, genes that are unlikely to be significantly differentially expressed due to very low counts across all samples are excluded before statistical tests. This is done to reduce the impact of multiple hypothesis correction. As a result, the 'between comparison' result table may display fewer genes than the total number of genes detected in the experiment. Please note that missing genes in the 'between comparison' table represent genes with insufficient data for reliable testing and thus are filtered out.
For whoever picks this up I would suggest to show this language, if possible, in the table where it says "No Data" in the following situation: 1) a between group comparison is run 2) and then the user searches for a gene and there are no results
I think the above is the primary scenario when a user will be wondering "why can't I see the results for gene x?". It may also make sense to show the above message when batch DE is run for between-group comparisons.
saracastel
Background In the differential expression tab in Data Exploration, there's a need for better clarification regarding the differences observed in the number of genes displayed in the results, between the within-group and between-group comparisons.
Solution Add a tooltip that provides this necessary information. The tooltip could be displayed when the user hovers over the 'Gene' column in the results table, but feel free to put it in a different place if you have a different idea. The tooltip should contain the following text:
"For between-group comparisons, genes that are unlikely to be significantly differentially expressed due to very low counts across all samples are excluded before statistical tests. This is done to reduce the impact of multiple hypothesis correction. As a result, the 'between comparison' result table may display fewer genes than the total number of genes detected in the experiment. Please note that missing genes in the 'between comparison' table represent genes with insufficient data for reliable testing and thus are filtered out."