Sequencing control for ChIP-Seq

[MayaAM]

Hello,

I have a question regarding sequescing control for ChIP-Seq:

In the chip we sequence only the regions of the DNA that the protein was bind to. Therefore, the part of the DNA that we squence for the chip is small.
On the other hand, the control containes the whole DNA.
Therefore, using the same number of reads for both of them, will derive extremely low coverage in the control compering to the chip, and we can get false picks.
I thought that that it might be a good idea to sequence more the control.

What do you think?

Thanks,
Maya

[rory]

The real comparison of interest is not between coverage of bound sites vs. coverage of the control, but rather having comparable sequencing depth between the ChIP background and the control.

"In the chip we sequence only the regions of the DNA that the protein was bind to" is not what happens in real ChIPs. Generally the majority of reads in the ChIP sample do not map to sites where the protein is bound -- often less than 10% of the reads end up around "real" binding sites. The ChIP only enriches the proportion of reads near binding sites, e.g. from 0.01% of the fragments to 10% of the fragments. Those 10% give much deeper coverage to the 0.01% of the DNA that is bound, while the other 90% of the reads are spread across the rest of the genome, forming the background. The main goal of peak calling is to separate binding site reads from background reads.

A lot of what is accomplished by the control is to mask out false peaks in the background reads of the ChIP, which are sequenced to a similar depth as the control.

If you look at it this way, it could be argued that the control should be sequenced somewhat less deeply than the ChIP. Standard practice is to sequence the controls to the same read count as the ChIP. It is possible that there may be places where there is a peak in the control that is not large enough to mask a false peak in the ChIP. However if you over-sequence the control, you may mask out true peaks. In general, biologists hate using their sequencing resources on controls, so without a clear demonstration of why a more deeply sequenced control is required, the current standard is unlikely to change.

Cheers-
Rory