Should I use False Discovery Rate Analysis or Post-hoc Test on Normalized Data?

Hi,

I have used Diamond/MEGAN pipeline to analyse Shotgun Sequencing Data. I obtained the normalized counts for all the samples (comparative analysis) and used STAMP software for statistical analysis. If the data is already normalized, should I use any False Discovery Rate (FDR) test with ANOVA? I am confused that FDR may decrease the significance of differential taxa if I have already used normalized counts to remove systematic bias.

Any comments would be very helpful. Thanks.

Best regards,

Faizan Saleem