Questions about RNA-seq experiment design

[tineax]

I try to sequence the RNAs from 2 conditions, 5 replicates per condition, using multiplex Illumina Hiseq 2000. I would like to consult about experiment design. There are 3 plans. Any suggestions are greatly appreciated, especially for the third one.

First, 5 samples are put in one sequencing lane (2 for condition A, 3 for condition B, and vice versa). This will be asymmetric.

Second, randomly choose 4 samples in one lane (2 for condition A, 2 for condition B). This reduce the replicates but increase the reads number for each sample.

Third, keep 5 reps each and pool all 10 and then run half a pool each on two lanes. Each lane would generate 10 separate fastq files and one could then concatenate the 2 fastq files (from the 2 lanes) for the same samples to get more reads for each sample, ending up with a total of 10 fastq files. We are not sure if there are other considerations in this scenario.

[Bukowski]

We do all our RNA-Seq through option 3.

Obviously this is not always possible when you have hundreds of samples, and then the randomised/mixed approach is fine.

But with 10 samples, just index them all separately, pool and sequence down two lanes and combine the Fastq.

[james hadfield]

Option 3 every time.

You can multiplex up to 96 samples in one pool on HiSeq. I always recommend pooling the whole experiment if possible, and if not consulting with a statistician beforehand to get a controlled randomisation.

A MiSeq QC can help chekc your pool is balanced. However for your experiment I'd suggest getting the seqeuncing done without MiSeq as a single-end 50bp read (what we'd use for DGE) is cheaper thana MISeq QC run in many labs!

[westerman]

A problem we have is balancing the number of reads between samples. As James suggested using a MiSeq is a good way to do figure out the balance before running the HiSeq. But even a better method (if you have the time) is to run a pooled sample in a single HiSeq lane and use the information from that lane to balance a second run. You'll get much more data that way.

[tineax]

Thank you very much. I really appreciate your help.