search

zhangyuqing / ComBat-seq

Batch effect adjustment based on negative binomial regression for RNA sequencing count data
154 stars 39 forks source link

ComBat-seq + DESeq2 #7

Closed 14lwt1 closed 3 years ago

14lwt1 commented 4 years ago

Hi, I am wondering if it is valid to input batch adjusted counts from ComBat-seq into DESeq2's standard workflow. Thank you.

zhangyuqing commented 4 years ago

Hi @14lwt1, yes, it is valid to input combat-seq adjusted data into DESeq2.

alibumizj commented 4 years ago

Hi, DESeq2 workflow has considered batch effects in group design ,should we need remove batch effects before DESeq2 difference analysis? Thank you Best wishes

PoisonAlien commented 4 years ago

Same question here. Should we remove Batch as a covariate?

adjusted <- ComBat_seq(counts = as.matrix(counts), batch = pdata$Batch)

dds = DESeq2::DESeqDataSetFromMatrix(countData = adjusted, colData = pdata, design = ~ Batch + Condition)
OR
dds = DESeq2::DESeqDataSetFromMatrix(countData = adjusted, colData = pdata, design = ~ Condition)
zhangyuqing commented 3 years ago

@alibumizj @PoisonAlien When considering batch effects in group design, it takes into account the mean differences across batch, not necessarily the variance differences. ComBat-Seq is designed to address both mean and variance batch effects. In theory, no, you do not need to include batch as a covariate any more. However, you can always try both and evaluate the results.