ID translation and remove duplicated genes which different probes are mapped are routine procedure in differential expression analysis using GEO series. However, when I do differential expression first then translate ID and remove duplicated genes, the number of DEG exceeds the former. Why?
That's probably because of multiple testing. If you perform DE analysis BEFORE removing duplicates, the number of genes used in DE analysis is larger than that of DE analysis AFTER removing duplicates, and it will end up with lower FDRs because there are more genes. And lower FDR can be leaded to more DE genes.