Deal with RNA spike-in in single cell analysis

[Bidfudge]

Hi everyone,

I'm working with single cell RNAsequencing data obtained with the fluidigm technology. In the protocol, we chose the 96 well-plate which allows us to get full length mRNAseq from the cell.
We followed the protocol from fluidigm and It doesn't provide ERCC spike-in to normalize the data but Ambion AM1780 spike to control the efficiency of the reaction in the C1. So we add 3 spikes in reaction.

We received the data from the sequencing and I'm starting to analyze them. I'm a little bit stuck on how to deal with the spike.

I have a quiescent population of cells, which do not express a lot of gene (2000 detected genes), and the percentage of Ambion spike in these libraries is closed to 50%.
In an other population of cells, they express close to 8000 gènes, and the representativity of spike are close to 5-10%.

I started to analyzed the data with Seurat, and in the function Seurat::NormalizeData(), it takes the size of the library to make the normalization.

As the representativity of the spike is not the same in all the library, I obtain really different results depending on the fact that I include the spike or not in the beginning of the analyses.

Does anybody get the same problem with this technologie, and know If I need to include the spike-in in the count-table to start the downstream analyses??

Thanks a lot,
Nicolas