MiSeq vs RNAseq

[memento]

I have a technical question - why longer PE reads are not considered better for RNAseq. Since MiSeq can produce ~7-8Gbp, 25million 250 PE reads should equal 125million 50 PE reads....whats the problem with longer reads and mRNA ?
Thanks!

[scbaker]

Originally Posted by memento

I have a technical question - why longer PE reads are not considered better for RNAseq. Since MiSeq can produce ~7-8Gbp, 25million 250 PE reads should equal 125million 50 PE reads....whats the problem with longer reads and mRNA ?
Thanks!

First, I'm not sure the MiSeq has really hit the 25M read mark yet. Their spec is for 15M reads, and that fits with getting ~8Gb with 2X250 reads. Having said that...

There's certainly nothing wrong with longer reads for RNA-Seq. Longer reads are generally better for alignment (being more likely to find a unique hit) and they are especially better when trying to sequencing through exon-exon boundaries.

However, you can’t easily equate 15M 2x250 reads with 75M 2x50 reads (or 150M 1x50 reads). It’s true that they’ll all produce the same number of total bases, but for RNA-Seq the number of independent ‘counts’ is a critical factor. Whether your read length is 50b or 2x250b, the MiSeq only generates 15M ‘counts’ and that’s what is going to define the level of sensitivity and dynamic range of the experiment.

[mcnelson.phd]

For a lot of people doing RNASeq, they have a reference genome that they can easily map their reads to, so the read length isn't a big issue for them compared to having coverage. Generally this is because of inefficient rRNA removal or the need to examine low abundance transcripts. So, even though the MiSeq offers longer reads, the HiSeq is vastly superior on a cost per sequence basis, so most researchers will spend just a little more $$ to sequence their samples on a HiSeq.

For those without a reference who need to do some de novo assembly, the longer read lengths will technically be better, but generally coverage is still king so the HiSeq rules the field there as well.

Saying that, I've done some single bacterial strain transcriptomes on our MiSeq to examine how a mutation effects various aspects of the physiology. Our reads were >80% ribosomal, so even though I had ~15M reads there wasn't a lot of useful information.