Dear fellow DEseq2 people,

I have a question concerning the usage of DEseq after batch correction.

I am looking at different samples from different timepoints and as we collected two samples at a different time in a different lab, I wanted to correct for batch effect. However I am not sure for the design. Should we still include the batches as part of the design even when we do combatseq before? Or should a better model just be design= ~timepoint ?

#the batch effect is on day_4 and day5_4
batches <- c(rep(1,3),2,rep(1,3),2,rep(1,6))

#we have 4 different groups we want to study
groups <- c(rep(1,4),rep(2,4),rep(3,3),rep(4,3))

corrected_data <- ComBat_seq(counts = countmatrix, batch = batches, group = groups)

dds <- DESeqDataSetFromMatrix(countData = corrected_data,
                                         colData = coldata_batch, 
                                         design = ~ batches + timepoint)

Also afterwards for calling DE genes:

dds1 <- DESeq(dds, test = "LRT",reduced = ~ batches)

Should the batch effect be part of the reduced model when using the LRT or should I test against ~ 1

Thanks in advance for any helpful comments!

Johannes

Either use raw counts with batch in design as in the vignette or the combatseq 'adjusted' raw counts without batch in the design as batch is already regressed in these. A combination makes no sense.