Enable users to filter genes by differential expression metrics (SCP-5090)

[eweitz]

This enables users to find genes by significance and size of differential expression.

Previously (#1768), genes could only be sorted by these differential expression metrics. Now, DE genes can also be filtered to exclude results that aren't significant, or have low levels or differential expression, or both. New range slider facets let users filter by those metrics. This satisfies a stakeholder request to enable ranking genes by log₂(fold change) after applying a <= 0.05 threshold for adjusted p-value.

Here's how it looks!

Range_filter_slider_facet_for_DE_genes_table__SCP_2023-07-13.mov

In addition to layout shown above, the facets are also somewhat responsive to different viewport widths. Analytics on changes to the range facets are logged to Mixpanel.

Test

Enhanced automated tests verify some range facet functionality. To manually test:

• Go to "Human milk" DE pilot study
• Select annotation "cell_type__ontology_label"
• Click "Differential expression" button
• Change range slider facets for "log₂(fold change)" and "Adj. p-value" to filter for different ranges
• Confirm DE table changes, either in DE genes shown on page in view or in page count at bottom
• Click checkbox beside range facet, confirm it makes the facet inactive -- i.e. doesn't apply any filters in that facet
• If you have author DE populated for the "General_Celltype" annotation:
• Select annotation "General_Celltype"
• Confirm range facet "Adj. p-value" does not appear, and that "q-value" does instead
• Change facets, confirm shown DE genes satisfy them

This satisfies SCP-5090.

[bistline]

Really great! Looking forward to this being available.

Side note that doesn't warrant action here: in the case were a user selects a comparison in the differential expression picker, and the applied filters result in no matches, should we consider some kind of notification to that effect? Otherwise, it is indistinguishable from being unable to retrieve results from the bucket.

[bistline]

I'm wondering if facets is potentially confusing here - we have plans to add actual search facets to many of our Explore tab plots this quarter, and this might get overloaded.

[eweitz]

Agreed, this confusion seems inherent in having two search UIs in view at the same time. The search UI in the DE panel queries a corpus of genes, and the search UI for the plots queries a corpus of cells.

To mitigate potential future confusion, in a489534 I just prefixed a lot these DE search variables with "de", so the cell search variables can be unambiguously named without a prefix.

[eweitz]

in the case were a user selects a comparison in the differential expression picker, and the applied filters result in no matches, should we consider some kind of notification to that effect?

Yes, that makes sense! For our reference, we just noted we'll do this as part of SCP-5227.

[ehanna4]

Looks good! confirmed works locally

[eweitz]

The build failure seems like a false positive, since the cause was a Rails test failure but not Rails code changed.

### FAILURES and ERRORS ###
*** Yarn test failures ***

*** Rails test failures ***

  FireCloudClientTest#test_delete_workspace:
  RuntimeError: Wrapped undumpable exception for: RestClient::InternalServerError: 500 Internal Server Error
Exiting with code: 1

The specific failure line essentially uninformative:

fire_cloud_client_test.rb:148 test_delete_workspace completed (20.843 sec)
E

I saw no test-logs output, I suspect because this branch doesn't yet include #1820. However, the false-positive rationale seems sufficient to merge now.