Hello,

I have counts from Salmon that I have imported using Tximport for WT and KO samples. I want to use ERCC spike-in data for normalizing these counts. To do this, my strategy involves estimating size factors using only the spike-in data (function: Deseq2::estimateSizefactors) and using those normalize my counts (Mus musculus transcripts).

What I have done so far: step1: I used Salmon for quantification of both Mus Musculus and ERCC RNA simultaneously (using a concatenated cDNA file). I imported these counts into R using the package "Tximport".

step2: I made two ddsTxi objects, one with ERCC spike-in tx2gene file and the other with mouse Ensembl tx2gene file.

step3: I then used the Deseq2::estimateSizeFactors with the Spike-in_ddsTxi object to get size factors. However, I am unable to get sample-wise size factors:

sizeFactors(Spike-in_ddsTxi) = null

I am aware that normalizationFactors(Spike-in_ddsTxi) gives me a matrix but I am not sure how to use this for normalization.

Can I please get advice on the following: Question1: Is my method above the correct way of going about normalization with spike-in data? Question2: If the answer to question 1 is no, then what method should I use? Question3: What is the difference between using estimateSizeFactors with Salmon data imported using Tximport vs some other count data e.g. from featureCounts? Question4: Lastly, what is the "control genes" parameter in estimateSizeFactors function? Is that what I am supposed to use?

Thank you.

Basically, just import one matrix as a dds and use controlGenes to specify the spike ins. This will take care of all the details for you. It will estimate size factors over the spike ins and then combine the size factor and average transcript length normalization to produce normalization factors.