Hello, I am trying to combine multiple bulk RNA Seq datasets which contain multiple conditions in all of them. I obtained the counts matrix using featurecounts and I am not sure if I should normalize each dataset and then combine them using ComBat or Limma or combine them first and then normalize and log transform. Additionally, I would like to perform differential expression analysis and gene set enrichment analysis on the dataset, therefore would be happy to know the best course of action here.

This question is off-topic for this site. You might try over on biostars.org instead.

Hello, I would recommend you use BatchQC (bioconductor package) to evaluate the extent to which batch effects impact your data. Not all batches of data require correction, and the less you need to do the better. Can you ignore the batch effect? Can you merely include batch as a covariate in your model? Or do you need to apply ComBat to your log counts per million or ComBat-Seq to your counts data? It depends on the dataset and your batch effects which is the best strategy--.