compute DE+ and DE- separately

scverse / scvi-tools

Deep probabilistic analysis of single-cell and spatial omics data

http://scvi-tools.org/

BSD 3-Clause "New" or "Revised" License

1.16k stars 341 forks source link

compute DE+ and DE- separately #2826

Open LevyNat opened 3 weeks ago

canergen commented 3 weeks ago

Fixed the wrong behavior that we have seen in different models (high significance despite low mean LFC but high standard deviation). It still requires a release note and it would be good to check in our DE tutorial whether things are changing there. I’m not sure that the Bayes factor computation is correct @PierreBoyeau?

PierreBoyeau commented 3 weeks ago

Hi, LGTM. I am unsure about the Bayes factor here, since we don't have two, but three competing models now(equally expressed, upreg, and downreg). I would be in favor of not computing BFs in this case.

canergen commented 3 weeks ago

I guess it's fine to remove. I would assume we do two tests: upregulated vs rest and downregulated vs rest. We output only the winner of the test for upregulated/downregulated. I guess the corresponding Bayes factor for the returned FDR is then: upregulated/(equally+downregulated) or downregulated/(equally+upregulated).