Diving in a bit deeper a few things have come to light.
EBSeq apparenlty creates more false positives than traditional deseq or EDGEr.
https://pmc.ncbi.nlm.nih.gov/articles/PMC6357553/
While all methods were highly accurate, half of the methods suffered from a low sensitivity (correctly identifying DEGs). Consequently, these methods had high false discovery rates (FDRs) and low precision as reflected by the overall measure of the F1 score (combined score of precision and sensitivity) (Table 2). In more detail, EBSeqHMM [25] identified more false positives (FPs) than true positives (TPs).
Surprisingly, TC tools were outperformed by the classical pairwise comparison approach on short time series (<8 time points) in terms of overall performance and robustness to noise, mostly because of high number of false positives, with the exception of ImpulseDE2.
In light of this, we can use our existing runs "as corroborating evidence" to the Deseq2 data. It is not worth diving much deeper than that.
Other issue is that EBseqHMM is no longer supported by the author, has been dropped from Bioconductor. So now it is a major hassle to run the tool. Might be worth exploring ImplulseDE2 instead in the future.
So closing this ticket.
Task
Major
Diving in a bit deeper a few things have come to light.
EBSeq apparenlty creates more false positives than traditional deseq or EDGEr.
https://pmc.ncbi.nlm.nih.gov/articles/PMC6357553/
While all methods were highly accurate, half of the methods suffered from a low sensitivity (correctly identifying DEGs). Consequently, these methods had high false discovery rates (FDRs) and low precision as reflected by the overall measure of the F1 score (combined score of precision and sensitivity) (Table 2). In more detail, EBSeqHMM [25] identified more false positives (FPs) than true positives (TPs).
Surprisingly, TC tools were outperformed by the classical pairwise comparison approach on short time series (<8 time points) in terms of overall performance and robustness to noise, mostly because of high number of false positives, with the exception of ImpulseDE2.
In light of this, we can use our existing runs "as corroborating evidence" to the Deseq2 data. It is not worth diving much deeper than that.
Other issue is that EBseqHMM is no longer supported by the author, has been dropped from Bioconductor. So now it is a major hassle to run the tool. Might be worth exploring ImplulseDE2 instead in the future.
So closing this ticket.