chanzuckerberg / cellxgene

An interactive explorer for single-cell transcriptomics data
https://chanzuckerberg.github.io/cellxgene/
MIT License
631 stars 118 forks source link

differential expression with low number of genes #2597

Open hschult opened 1 year ago

hschult commented 1 year ago

Dear cellxgene team, first of all, thank you for developing and maintaining such a nice tool!

I am trying to analyze a citeseq dataset with cellxgene. The difference here is that instead of cells vs. genes, I have cells vs. a low number of antibodies. In my case 42. Anyway, whenever I try to do a differential analysis on the data I get the following popup: grafik The number of selected cells doesn't seem to make a difference and everything else works fine.

With a bit of trial and error on a different dataset I could confirm that differential expression only seems to work for > 50 genes. Also, I found this pull request #2130 which might be related.

Versions:

Best, Hendrik

MaximilianLombardo commented 1 year ago

Hey Hendrik - happy holidays - the CELLxGENE is taking a holiday break this week. I'll be back January 9th to address this (unless someone beats me to it :) ). Happy new year!

MaximilianLombardo commented 1 year ago

Hey Hendrik - happy new year and thanks for your patience on this issue.

The PR you linked does seem to be relevant - would be able to pad your dataset with a few extra features (aka so that it has more than 50 variables i.e., add some placeholder gene names and set expression values all to 0) to confirm that having less than 50 features in the dataset is what causes the error to be thrown?

I'm also following up with the team on this.

hschult commented 1 year ago

Hi Maximilian, thank you for looking into it. I hope you had a nice holiday.

I can confirm that adding placeholder features to the dataset resolves the error. I have used this approach as a workaround now.