Will samtools merge influence FPKM?

[EricHole]

Hi All,
I have replicateA.bam and replicateB.bam. Then I use samtools merge replicateA.bam and replicateB.bam. After that, I use cufflink to do the fpkm calculation. Will Fpkm from this result double compared to caluculatation one by one? Maybe because the later one have less reads? The merge one have double reads?

Thank you for your help.

[Nicolas]

Do NOT merge replicates! You're losing one useful piece of information (ie that they're not the same experiment, they are replicates!)
Instead, use cufflinks and indicates the two BAM files as replicates, separated by commas.(replicateA.bam,replicateB.bam).

[EricHole]

Do NOT merge replicates! You're losing one useful piece of information (ie that they're not the same experiment, they are replicates!)
Instead, use cufflinks and indicates the two BAM files as replicates, separated by commas.(replicateA.bam,replicateB.bam).

Thank you!!! Could you also explain will the result double?

[Nicolas]

not necessary. Cufflinks do some magic with the reads leading to FPKM and it's hard to follow precisely all the normalisation step (some people may disagree). I would imagine that the results will be closer to an "average" than to a sum, but for sure it will be more robust this way.
I don't use Cufflinks, prefer to stick to raw counts (obtained with htseq-count), and DESeq / edgeR / limma for the differential expression analyses.

[EricHole]

Great thanks!!