Hi all,

I am aiming to merge my dataset with other available online. Data will be SCTv2 normalised, and visual integration will be done by Harmony.

For DGE analysis, I would like to perform pseudo-bulk comparisons (either via EdgeR or DESeq2). I have got two questions:

1) Should I aggregate raw counts or PrepSCTFindmarker counts 2) What's the best way to account for batch effect across datasets?

Thanks

Hi KuriGura

Regarding DESeq2:

1) In my opinion the safest would be to use counts from RNA assay as input for DESeq2. DESeq2 has pretty robust normalization methods and will account for sequencing depth. Also, i think going for SCTv2 corrected counts should not violate any underlying assumptions but lets see wether developers will comment. For future reference, it is always useful to use tags for software in question (DESeq2 and EdgeR in this case - you only used scRNAseq tag)

2) In DESeq2 you can account for batch with appropriate formula. For example:

`dds <- DESeqDataSetFromMatrix(bulk_clus_counts, colData = sample_meta, design = ~ batch + condition)`

where batch is column with your batch info and condition is defining groups which you want to compare.