View Single Post
Old 06-23-2016, 09:58 PM   #5
Brian Bushnell
Super Moderator
Location: Walnut Creek, CA

Join Date: Jan 2014
Posts: 2,707

I have not noticed problem A (it is something I may look into... but it sounds like a conflated library-creation issue) and have no experience with problem B. But I will add that NextSeq has substantially lower quality than HiSeq, in my experience, so I suggest you carefully consider whether the decreased sequencing cost on a NextSeq is justified. Normal RNA-seq quantification is quite tolerant of error. Small RNAs... well, microRNAs are so short that you normally want to map them allowing zero mismatches. So, any errors are bad, and biased errors are extremely bad. But since they are so short, errors decrease the signal-to-noise ratio much more than they would for longer sequences, even if you allow zero mismatches, because with a 17-mer (for example) you can easily get coincidental perfect matches from reads with errors.

Unless you do not care about very short RNAs (under 50bp or so), I suggest you use the highest-quality platform available, which is MiSeq or HiSeq 2500. You can improve things a lot by merging overlapping reads, too, so aim for read lengths that substantially overlap (by at least 50% of the read length, or ideally 100%, if you want to minimize errors) in the size range of interest.
Brian Bushnell is offline   Reply With Quote