Should I do RNA seq read counts normalization(CPM) before filtering protein coding genes?

Question

[dustin3141]

I have RNA seq data with 29 samples and 60483 genes, but I only want to focus on protein coding gene. I filter protein coding gene and filter low count then do CPM normaliztion. Is this method reasonable?

[ATpoint]

Without knowing more details I would leave all genes in for normalization. Be sure to use established normalization methods such as TMM in edgeR (cpm function) rather than naive per-million scaling. After this you can filter. What is the final goal? Differential analysis?