Cuffdiff vastly different results

[jfreimer]

Hi,
I'm new to RNA seq analysis and have been having some weird results. I have cell type A that I'm differentiating into cell type B. I have grown these cell types in two similar, but slightly different media conditions, which I'll just call 1 and 2. Since this is a preliminary experiment I was sharing a flowcell with other samples and do not have replicates for each condition.

I have mapped them all to the genome and have sorted bam files which I used as input for cuffdiff. I am just looking at differences in the reference transcripts and so I've been skipping cufflinks and using the reference annotation gtf file that you can download from Illumina/the cufflinks website.

When I compare cell type A to cell type B in media 1 there are 0 genes significantly different. I see the same thing in media 2. However, if I treat cell type A in media 1 and 2 as replicates and cell type B in media 1 and 2 as replicates I now see ~ 8700 genes significantly different between the two cell types. Is this just because the cell types are too different to call significantly different genes without replicates?

Here is what my run command looks like for the single samples:
cuffdiff -b /N/dc/scratch/tg-jfreimer/UCSC_mm10_bowtie/Sequence/WholeGenomeFasta/genome.fa -u --library-type fr-firststrand --dispersion-method blind /N/dc/scratch/tg-jfreimer/UCSC_mm10_bowtie/Annotation/Genes/genes.gtf A-1.sorted.bam B-1.sorted.bam

and for replicates
cuffdiff -b /N/dc/scratch/tg-jfreimer/UCSC_mm10_bowtie/Sequence/WholeGenomeFasta/genome.fa -u --library-type fr-firststrand /N/dc/scratch/tg-jfreimer/UCSC_mm10_bowtie/Annotation/Genes/genes.gtf A-1.sorted.bam,A-2.sorted.bam B-1.sorted.bam,B-2.sorted.bam

Thanks for any help!

[dpryan]

Regardless of how similar or dissimilar the conditions are, you'll always need replicates to get meaningful results. Next time, just get a lane for yourself (they're relatively cheap) and stick a few replicates on it.