Johannes Ostner

Hi @Lopiniatre! We are aware of all these warnings, and they should not affect the results. A quick breakdown: - The first two tensorflow warnings refer to GPU optimization which...

Hi Koen, `scCODA` uses a thresholding on the posterior inclusion probability of each cell type to decide whether they are differentially abundant. The nominal FDR approximately determines, where this decision...

One additional remark: A FDR of 0.4 is quite high - meaning that you expect 40% of your discovered effects to be false positives. If this was just to test...

Thanks @koenvandenberge for pointing these things out! Yes, a larger number of posterior samples can decrease the between-run variability, but only up to a certain point. We observed some fluctuation...

Hi @koenvandenberge! Thanks for taking the time to prepare a mock dataset. I've looked at the data and your analysis - here's what's going on: Looking at the relative abundances...

I tried running the code you posted again and still get the same result as before. In the screenshot you just posted, it looks like only the result of the...

Hi, thanks for this pull request! I am unsure though, whether this change is necessary. As far as I know, it is also possible to get consistent results by running...

Hi @c-westhoven, when we updated scCODA recently to support the latest tensorflow (>=2.8) and tensorflow-probability (>=0.16) versions, we missed updating the readme. The correct version requirements should be displayed there...

Hi! Unfortunately, mixed effect models are not implemented for scCODA. Some alternatives would be to encode the donor info as a categorical variable or to do a donor-wise analysis, if...

Hi @dawe! We've regularly faced similar issues in the past, which were usually related to breaking changes in tensorflow or tensorflow-probability. Therefore, we recently re-implemented scCODA as part of the...