Validation of Methods for Low Volume RNAseq

BioRxiv link:


So this is mainly a set of control experiments that are not that exciting in themselves, but that I thought would be useful to put out there.  The question boils down to: "how little RNA does one really need to make a good differential-expression type RNAseq Library?"  There was plenty of anecodtal evidence from our lab that the official Illumina protocol was playing it safe, but no one had really tested it out. 


Some key punchlines:

  • To get really reliable library preparation using standard TruSeq reagents, I need about 60-70 ng of total RNA. This is almost half of what the manufacturar says you need, but not really small enough to use for slicing, at least not without adding carrier
  • There are "single-cell" protocols out there, and while they give almost uniformly lower quality data, even from the same number of reads, they can work with orders of magnitude less sample, which makes it worth it.
  • The expression values you get from these protocols are not directly comparable with each other, or with the standard protocol. But, and this is a big "but", the data are linear, so if there is twice as much of gene A in sample 1 vs sample 2, the expression value you get out is twice as big. This means we can use them for making spatial gene expression atlases.
  • You can dramatically drive down costs by just using less reagent per library.  This works to the point where sequencing and your own time become the major cost drivers.