Using both spike in and TMM normalizations in ChIP-seq samples
Entering edit mode
Last seen 23 months ago


We have performed a ChIP-seq experiment in S. cerevisiae with two conditions (control and mutant) using S.pombe as a spike-in (because the antibody works in both organisms). The reason because we add spike-in is that the antibody is not very efficient so we wanted to normalise the immunoprecipitation efficiency between samples in some way. But we would like to add another normalization step as TMM because the library size is very different between conditions (mutant condition accumulate most of the reads in repetitive regions that we are no able to map) and we expect a small portion of regions to be differentially enriched.

Does anyone knows how should we proceed?

Thanks in advance

ChIPSeq SpikeIn edgeR Normalization • 1.0k views
Entering edit mode
Last seen 52 minutes ago
WEHI, Melbourne, Australia

It is impossible to use both spike-in or TMM normalization. You must do one or the other. They replace each other, they cannot be combined.

TMM normalization is independent of whether the library sizes are different or not. edgeR always normalizes for library size and that is separate from TMM normalization.

I would recommend spike-in normalization only when there are likely to be massive global changes in binding, i.e., when more than half of regions are differential in the same direction. If you expect only a small portion of regions to be differentially enriched, then spike-in normalization is unnecessary and highly counter-productive and will simply add noise to your experiment.

I don't follow the background you give. You say that there is massive change of where the reads are mapped between mutant and control but you also say that only a small portion of regions to be differentially enriched. How can that be true? If it was true, then why would you not simply use TMM, which is designed for exactly that scenario?

Entering edit mode

First of all, thank you so much for your reply. It was really useful!

The change of where the reads are mapped is massive but also very located in certain regions (rDNA, tRNA, snRNA,...). I was not aware that edgeR always normalize for library size, then my problem is solved I think.

Taking into account all this new information, I think TMM is the method that suits better my experiment.

Thanks again!


Login before adding your answer.

Traffic: 610 users visited in the last hour
Help About
Access RSS

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.

Powered by the version 2.3.6