DESeq in libraries with spike-in samples

[frymor]

Hi all,

in our experiment we have control triplicates as well as two different knockouts. for each of the conditions the total RNA as well as two different fractions of polysomal (P) and non-polysomal (NP) RNAs were sequenced.

for the detection of translation efficiency the biologist added Spike-In controls into the P and NP fractions, but not to the controls.

As far as I understand it, DESeq uses the estimation of the effective library size to calculate the differentially regulated genes differentially regulated.

In the libraries with the Spike-Ins I have an enhanced library size of reads with no relation to the reference genome used in the analysis.

How do I run the DESeq analysis on these files?

I mapped the fastq files to the genomic sequence of the Spike-In and know how many reads were mapped to it.

Do I need to somehow normalize my libraries with the Spike-Ins to the number of reads mapped to this genomic sequences?

When we sequences 20 million reads, we get a lot more reads from the organism of interest, when we don't have the Spike-In in the probe. On the other hand, When I map the fastq files to the reference genome, the reads from the Spike-In controls won't map and therefore won't be in the bam files.

DESeq normalize anyhow according to the library size. Does it means, that I don't need to worry about these differences?

Thanks in advance.

Assa

[jparsons]

Assa,

You should be concerned with the effect of the differential spike-in between the treatment and control samples. This effect may be quite minor or more serious, depending on the percentage of spike to total in your sample.

Thankfully, you have the data you need. Feed DESeq the count data with the spike-in excluded (mapped only to the genomic sequence). It will handle the library size normalization correctly at this point.

To give yourself some peace of mind, you can calculate the sizeFactors both ways. You should expect the P+spike factor to be lower than the P-spike factor.

[frymor]

The problem I have is, that there are no Spike-In in the control samples, but only in the two treatment conditions.

I my mapping I used the genome without the Spike-In sequences. So I guess, I don't really have to woory about "contamination" of my library sizes with residues from the Spike-Ins.

Thanks for the advice

Assa