ChIP-Seq annotation question: Should gene annotation based on the nearest TSS or gene body?

Question

[shimengliu11]

Hi,

I am currently doing gene annotation for ChIP-Seq. I am wondering which annotation method should I use. Should I annotate the binding region based on the nearest gene body or the nearest transcription starting site?

For instance: Gene A is a 100kb long. Gene B is 10kb down stream of Gene A. I found a binding site at the end of Gene A (eg. 90kb downstream of Gene A's TSS and 20kb upstream of Gene B's TSS). When investigating which gene the binding region will influence, should I annotate the binding region to Gene A or Gene B?

Thank you in advance for any help.

Answer 1

[GouthamAtla]

Usually, its very hard to decide which gene it influence. It could influence Gene C which may be farther away than Gene A or Gene B. Best bet would be, check if both genes are expressed in the cell-type of your interest (if one of them is not expressed, you could eliminate that) , check if both genes are in same TAD (if one of them is in different TAD, you could eliminate). Check if you have "Hi-C" like data for that cell-type to know exact target.

Or simply use GREAT to find likely targets.

Answer 2

[jared.andrews07]

This is a tricky question, as regulatory elements can impact genes that are several 100kb away. Higher confidence annotations really require more information, such as correlations with gene expression, chromatin interaction information, higher order genome structure (looping, TADs), and more. Most packages meant for ChIP annotation (clusterProfiler, rGREAT) go by distance to TSS. GREAT has different association rules and settings, with the default allowing for annotations to multiple genes.

It's fine to annotate this way depending on your goals, but you need to take it with a grain of salt, as the associations will likely have a high degree of false positives/negatives.