Hi everyone,

I have RNASeq Data, with 4 conditions, 2 replicates each (very far from ideal, I know). By looking at my data through heatmap and PCA, I see that there are relevant differences within conditions, among replicates.

To further analyze this dataset, I got the recommendation of using betaPrior=TRUE instead of the default FALSE. What I would like to know is if there's any scenario where using this option is more recommendable than following the standard pipeline (betaPrior=FALSE w/ lfcshrink).

Thanks!

Not really a use case. It’s obsolete except we kept it to not break old code that used it. The apeglm paper shows that the new heavy tailed methods (apeglm and ashr) outperform the Normal prior pretty much in every case.