The theoretical logFoldchanges from SymSim compared to the ones reported by other methods, e.g. scanpy, on the true_counts are rather different. I expected that there is at least some correlation, because for sure the values between different methods differ, but it is calculated on the same data without any further noise.
Unfortunately it is not the case and both methods return quite different values. Because we found that scanpy's logfoldchanges are correlated with those from other DE testing frameworks (tested on other datasets), I assume that the underlying cause is in getDEgenes. Maybe there is just some usage mistake, where you can help and clarify the difference. Thanks in advance for your help.
The theoretical logFoldchanges from SymSim compared to the ones reported by other methods, e.g. scanpy, on the
true_counts
are rather different. I expected that there is at least some correlation, because for sure the values between different methods differ, but it is calculated on the same data without any further noise. Unfortunately it is not the case and both methods return quite different values. Because we found that scanpy's logfoldchanges are correlated with those from other DE testing frameworks (tested on other datasets), I assume that the underlying cause is ingetDEgenes
. Maybe there is just some usage mistake, where you can help and clarify the difference. Thanks in advance for your help.Code to reproduce:
And scanpy for comparison:
Versions: