When to merge specify replicates

[nr23]

Hi,

I've got RNA-Seq data (pe 100bp) from several biological replicates and 11 different conditions and want to compare DE between conditions.

I'm using the tuxedo suite (tophat, cufflinks) and wanted to know if I should run both replicates through tophat at the same time (specified as reps) , and use the output for cufflinks, or separately, and specify them as replicates in cufflinks?

Additionally, the samples have been multiplexed (so each replicate is split over two lanes) - and I've currently merged the multiplexed reads for each replicate - do I lose power by doing this?

Any advice would be great!

Thanks,

N

[mbblack]

Keep your biological replicates separate, and treat them as replicates. Otherwise, there was little point in even running them as you will have completely obscured the biological variation amongst your samples.

Merging all your multiplexed lanes per sample prior to mapping them is fine and then you can map them all as the single sample they represent.

[mastal]

When to merge specify replicates

How many lanes were used in all? Were all of the samples run in the same two lanes?

You shouldn't combine biological replicates when running Tophat. You need the replicates in order to estimate the variance when calculating DE.