**10**wrote:

I have a ChIP-seq dataset with two conditions. For each condition I have two replicates of both input and the acutal ChIP experiment: 8 samples in total.

In R the design matrix would look like this:

myDesing <- data.frame(sample=1:8, data=c(rep('Input',4), rep('Chip',4)), condition=c('Wt','Kd'))

myDesing$condition <- factor(myDesing$condition, levels=c('Wt','Kd'))

myDesing$data <- factor(myDesing$data, levels=c('Input','Chip'))

> model.matrix(data=myDesing, ~data + condition)

(Intercept) dataChip conditionKd

1 1 0 0

2 1 0 1

3 1 0 0

4 1 0 1

5 1 1 0

6 1 1 1

7 1 1 0

8 1 1 1

I'm interested in how to handle the input data in regards to doing a differential binding analysis with edgeR, DESeq(2) or Voom-Limma.

Is it possible to incorporating the input data into the linear model?

My idea is to make a model that includes both the input and chip libraries and then test the interaction between chip and condition. More specifically for the R code example above this would correspond to testing the interaction dataChip::conditionKd. Does this make sense or does it violate some of the assumptions?

Does this not correspond to measuring the fold change of Input vs WT ( FCiwt ) and the fold change of Input vs KD (FCikd) and then comparing whether there is a difference between FCiwt and FCikd?

I have read this DESeq2 for ChIP-seq differential peaks which discuss whether one can subtract the input reads from before the DE analysis but it is a bit unsatisfactory as the discussion did not result in a good way to incorporate the input data.

In advance thanks for your help.

Kristoffer

**17k**• written 4 months ago by k.vitting.seerup •

**10**