Column/Sample names not in alphabetical order may cause DE conditions/results to flip

[lijingbu]

Great work for EBSeq, it makes more Biological sense in our project comparing to DESeq2 and EdgeR.
Just found out this issue below. Hope it can be tested.
It works well if samples (column names) are in order with control first/ on the left, like:

A1, A2, A3, B1, B2, B3 when using samples A as the control group, B as the testing group.

However, it seems to reorder the samples' test order, when
B1, B2, B3, A1, A2, A3 when using B as control, test sample A against B. I got reversed results, as if I am still comparing B against A. Thus my DE results were reversed, having up-expressed genes listed in the down expressed list and vice vasa.

My example results of in EBSeqOut. Notice my attention was to compare CB2 against CM2, not compare CM2 to CB2. But it assigned the Condition order and factor levels in reverse order.

$ConditionOrder
Condition1 Condition2
"CB2" "CM2"

$Conditions
[1] CM2 CM2 CM2 CB2 CB2 CB2
Levels: CB2 CM2

[lengning]

Hi Lijing,
Thanks for your interest! You may try to input the condition factor as
factor(Cond, levels=c("CM2", "CB2"))
Best,
Ning

[lijingbu]

Hi Ning,
Thanks for the reply.
It looks like the code was embedded in the EBTest.R script. In line 16 and 23. I simply run the function EBTest.R and got reversed factors.
Best,
Lijing