What is ERCC spike-in really saying about my sequencing run?

Question

[exin]

I've used ERCC spike-in to prep the library for my recent run of sc-RNA seq.

Looking at the plot of ERCC read counts vs ERCC concentration, I know we should ideally expect to see a nice correlation with R2 =1, for every sample in the library.

Looking at the sequences itself, QC and mapping %, I've had a bad run. Slopes of the ERCC plot are all over the board and anywhere from 0 to 1. What specifically does this ERCC plot really indicate about the sequencing run? Besides that "something went wrong" or that samples aren't evenly amplified? - How should I rectify this?

While we're at this, can someone explain or point me to a good resource on how to use ERCC counts to normalise my data?

Thank you!

Answer 1

[Kevin Blighe]

It's not new. These ERCC spike-ins don't behave as one would expect and I do not know how they found their way to the marketplace. The company selling them will have some pristine plots to show, of course. As such, they cannot be trusted, and I would not actually encourage you to use them for normalisation.

[exin]

Thank you for your insights Kevin!

[theodore]

Just my 2 cents. https://www.nature.com/articles/s41592-018-0265-y this is a good overview on the status of the spike-ins and if I am not mistaken it is referenced that the ERCC are not good for specific RNA-seq analysis that go beyond the microarray data. The reason is that ERCC were developed with the microarrays in mind.