Loading libraries in HIseq for RNAseq

[dhanapala]

• How many libraries can be loaded in each lane for getting good results with RNA seq

• When more libraries will be added whether the read quality will be same or the quality will decrease?

[peromhc]

are you asking about how many individually barcoded samples can be mixed and loaded in a single lane?

[KtWa8abug]

RNA-Seq: Number of Reads Needed

I would like to know the same. More specifically, how many reads are needed for accurate gene expression profiling of murine samples? One source says 8 million reads per sample are enough for mouse/human/rat samples but then goes on to call this amount of reads appropriate for a "pilot" study if funds are tight. Our funds are tight, but not so tight that we want to spend the money on data that we cannot publish or even trust our own interpretations for future direction.

Any feedback would be greatly appreciated!

[Jon_Keats]

To get good "gene expression" 30 million reads should do it, around 5-6 samples per lane. If you can get on a 50x50 flowcell that should make library prep the most expensive part of the experiment.

[TonyBrooks]

I haven't seen a consensus on how many reads you need for a mammalian RNA-Seq experiment. I've seen estimates of between 2 and 10 million reads to equal the sensitivity of the latest microarrays. I would definitely aim for more than that otherwise you defeat the purpose of NGS - you might as well run an array.
I would agree with Jon in saying that around 30 million is good amount, that means if you can quantify your libraries accurately (qPCR), pool to an equimolar amount and you hit the top spec of a HiSeq lane (187m reads), you should be able to run 6 samples in a lane (i.e. a simple 1 factor experiment with 3 biological replicates = 6 samples with ~30m reads each).
In my opinion, it's also better to run more replicates with fewer reads, than fewer replicates with more reads (i.e. 2 x 20m > 1 x 40m).

Also, it's actually better to run more indexed samples in a lane than fewer as the base calling algorithm can be dependent on cluster diversity. The more tags, the better it is (so I've been told by our Illumina FAS).