Hi all, this is Yet Another Spike-in Post, and I am sorry about it. So I was reading this paper: spikeIn with Drosophila
and I am troubled on the value of this approach. from the supplementary material and methods they just mixed D.m. cells with their mice cells:
50,000 S2 Drosophila cells were used as spike in control for the RNA-seq experiments.
A. Has anyone come across a method like that (that resembles the spike in of chip-seq experiments, but in this case chromatin from D.m. S2 cells is used) and do they have any comment on its value?
B. Also in the sup.methods they do not describe how this spike in has been used and also they use featurecounts and then RSEM or something else? Is that is uncommon and weird or they are doing something I do not comprehend?
We used the latest annotations obtained from Ensembl to build reference indexes for the STAR alignment. FeatureCounts (Liao et al., 2014) was used to count reads mapping to each gene. RSEM (Li and Dewey, 2011) was instead used to obtain FPKM (Fragments Per Kilobase of exon per Million fragments mapped). We analyzed differential gene expression levels with DESeq2 (Love et al., 2014), with the following model: design = ~condition, where condition indicates either ARID1A-WT or ARID1A-KD
Thank you all in advance
They simply add a constant amount of exogenous material to the library prep (or experiment, I did not read the paper, so my comment is general). That means this exogenous material will be the same / not change due to the experimental conditions and can therefore serve as a baseline for normalization or as a technical control during DE analysis. You can google "ERCC spike-in" and to get some background how this basically works. During quantification you then could add the spike sequences as additional "chromosomes" to your reference genome and as extra "genes" to the GTF file, or simply directly quantify against a reference transcriptome that contains these sequences, e.g. with tools like
salmon.thank you for the quick reply, they do not provide any detail on how the perform the normalization. Are they calculating scale factors or fit the mouse data to a distribution? since the cells undergo the same procedure won't they be affected by the biases introduced in the RNA purification steps? won't be a better alternative to spike purified D.m. RNA? even an mRNA ribo depleted one? any clues on the RSEM-feature counts description?