CHiP-Seq Questions

Question

[B.N.]

Hello,

New to chip-seq analysis so I apologize in advance for the questions. Currently using diffbind. I see that the interval number for one of the replicates out of the 3 is very high compared to the other 2 replicates. FRiP score across samples is also lower than 0.05. Number of counts is also a lot higher in 1 replicate in both groups. Should I discount this high interval sample or is it safe to use for diffbind analysis with the other replicates. After running diffbind once for a differential binding analysis with dba.analyze I see that I get very few significant differentially bound peaks.

Thanks

[jared.andrews07]

Sounds like you should do some QC of your ChIP-seq data first to ensure the assay actually worked.

Answer 1

[Rory Stark]

You may want to add Replicate to your design formula to model the batch effect (assuming the same replicate is the outlier in both conditions). For example:

myDB <- dba.contrast(myDB, design="~Replicate + Condition")

Library normalization (background windows or FULL_LIB) should be able to take care of the differences in read counts (sequencing depth).

As was suggested, the real issue is if you have adequate quality for the three replicates. You can check the correlation heatmap (dba.plotHeatmap()) and PCA plots (dba.plotPCA), after calling dba.count(), to see if the samples from different groups are clustering together.

[B.N.]

Thank you for the suggestions Dr. Stark.