I have 2 bacterial strains of the same species I want to perform differential expression analysis on. Both strains were grown twice in their respective media, so that RNA extraction resulted in 2 biological replicates per strain.
I chose to do my differential expression analysis with cufflinks, to find genes expressed at different levels, and also to calculate FPKM values for every replicate. My question related to what makes cuffdiff decide a gene is significantly differentially expressed?
Some times I see high log2 change values, but then large variability among replicates, and I understand how that makes the gene unreliable.
However, other times, the same log2 change can be seen with consistent replicates yet it is not deemed significant (even high pvalue). Does the magnitude of FPKM also play a role then?
Lastly, I have seen some papers where COGs are identified in transcripts and each COG category plotted as a percent of trnascriptome (based on FPKM counts). Is this a good idea? Especially since such studies usually also test for significance of difference among COG expressions.