I have always been confused by the order by which Normalization, (un)known batch correction, and QC should be performed (either within a single experiment or in meta-analyses). Could anyone help me or send me a reference article?
Typically you would do normalization first, followed by qc. Simply because QC (e.g. PCA) takes as input normalized values so nornalization must come first. QC would then tell whether any correction for known or unknown factors would be necessary or beneficial. Hence, typically it is normalization - qc - correction (if necessary). Even though in practice you might try different correction strategies and see how it affects results and whether results make sense, so a bit of going back and forth is not unusual.