Sample prep from different cells when amt of RNA is vastly different...?

[sjeschonek]

I'm doing an RNA-seq experiment comparing transcript expression across different stages of development.

My first stage contains about 60-fold less total RNA than the final stage. Most of the library prep protocols suggest starting with a certain concentration of RNA. If starting with 1 ug of total RNA input, the late stage sample would require only 1 ul of RNA, but the first stage would require 60 ul. I could dilute the highly concentrated RNA, or make the less abundant RNA more concentrated by using an increased sample size, but I'm worried about any bias this might introduce.

When comparing samples with vastly different amounts of RNA, what's the correct approach? Should all RNA just be diluted to a certain concentration? Or, should the starting RNA be based on volume (ie: 1ul of sample A, 1ul of sample B, even though total RNA concentrations would be very different).

Does a volume based approach more accurately capture the true picture? Does diluting to a specific total RNA concentration perhaps obscure less represented RNAs?

[cmbetts]

Most input related bias (Every library prep method is biased in one way or other, but usually consistently biased as long as you stay with the same protocol) in RNA-Seq isn't so much bias but noise. As you have fewer and fewer molecules going into the prep, the Poisson noise from the presence or absence of transcripts and library prep inefficiencies start to dominate the lower expressing genes and destroy your ability to get statistically relevant DE.
You'll be best off sticking with the same kit/protocol for all samples and trying to get as much RNA, up to the recommended amount, for each sample into the library prep. If you can't concentrate your early samples enough for this, it might be advisable to choose a protocol specifically intended for lower input. They generally have higher efficiency molecular biology that reduces the aforementioned noise issue.