I am going to meta-analyse four gene expression sets (they all have the same platform). Should I normalise each expression set before I meta analyse?
For intersample comparisons in RNA-seq, you generally want to normalize for library size. Either the TMM values from edgeR or the normalized counts from DESeq2 will account for this.
And to complement, before doing any intersample comparison, you could do some QC (e.g. PCA), after normalization but before any statistical comparisons. This would give you an idea if there any any other factors you should consider for your analysis.