I'm noticing that after running my BAM files through CuffDiff and CuffNorm that some genes that I know to be very highly expressed [Alb in liver should be 1M-2M reads; Gapdh should be fairly highly expressed] are showing zero [Alb] to under 20 reads [Gapdh]. I find the expected sized pile-ups when i check my BAM files on IGV. I'm certain this must be a parameter that can be adjusted in the Cuff* tools and is a consequence of many of these tools not really being written for full transcriptome RNA-Seq analysis. Any suggestions on how to fix this?
Here's my set up:
Mouse total RNA sequenced with Illumina HiSeq
Fastq to BAM files using TopHat2; defaults for all settings
BAM files straight to 'CuffDiff For CummeRbund' and 'CuffNorm' for tabular outputs; 3 Test replicates/5 Reference replicates
Using iGenome UCSC mm10 gtf file for both CuffDiff and CuffNorm