Normalisation of RNA-Seq data from cell fractions

[blancha]

Hi,

Does anyone know how to normalize RNA-Seq data from subcellular fractions?

I have RNA-Seq data from the nucleus, insoluble, cytosolic, and membrane fractions, as well as from the whole cell.

The same amount of RNA was submitted for sequencing. Spike-in reads were included before ribosomal depletion.

Normalisation was performed relative to the spike-in reads.

In my opinion, the comparison between the fractions normalized relative to the spike-in reads is invalid since it does not take into account the different amounts of RNA in the fractions.

Any comments, opinions or advice on this issue?

Thank you.

[jparsons]

I have done some empirical mixture validation where I show that spike-ins do an excellent job of determining the different amounts of RNA per mixture component, assuming that you spike into a known mass of total RNA.

The normalization calculation is a relatively simple one - I don't do any loess fitting or anything, i just determine a size factor as follows:

We calculate ρ as sample reads per microgram of total RNA divided by spike-in reads per microgram of spike-in RNA. (This is pre-supposing a similar number of total reads per sample - e.g. after an upper quartile normalization)

The ρ is then used as a scaling factor for each library.