Hi, we have RNAseq data that have been generated with different library preparation kits and need to be put together. I have learned that the best way to accomodate for this is including it as batch effect in the design formula of the DE packages (e.g. DESeq, edgeR). However, we would still need to batch correct the counts for additional analyses (e.g. PCA, clustering), and we were thinking on using limma::removeBatchEffect. Here comes the question: on which data can we use this function? Vst, rlog, log2CPM appear adequate, Is it so? Can we use it also on raw counts (e.g. those generated by RSEM), DESeq2 normalized counts, TPM, or the function expects a different kind of data? Do we need to log2 the counts/TPM before applying the function? Any advise against using counts treated with limma::removeBatchEffect for subsequent differential expression analysis using DEseq2 or edgeR? Sorry for these naive question and in case a duplicate question I was not able to find has been already answered before. Thank you in advance.
