advice on the size selection step in ChIP-seq protocol

[mboth]

Hi everyone,
This may be a really silly question, but why do we need to size-select the ChIP sample before PCR amplification? To me it makes more sense to keep as many molecules of the sample and amplify those, rather than limit the material by size selecting it.
I have tried out both, and although I get much more material of a tighter size range with a size selection step before PCR, I can also produce a library with a wider size range (but less DNA) without this step. At least this is what the bioanalyzer results tell me - I haven't sequenced them.
I am using 1xAMPure beads after adapter ligation, and expect them to get rid of my adapters.
Is there a flaw in my thinking? I hope somebody out there can share their experience with me.
Thanks!
Maike

[arolfe]

Don't consider this the authoritative answer, but I can think of two possible reasons:

1) Sequencing may work better with more uniform fragment sizes. I've heard varying claims about this on Illumina.

2) If you limit the length of fragments that you sequence, then your reads shouldn't be more than that distance away from the binding site. So only sequencing smaller fragments may make your binding site identification easier or more accurate because the reads will be clustered more tightly around the site.

[mboth]

Thanks arolfe, that does make sense.

[ETHANol]

Bridge PCR (cluster formation) works best when the DNA fragments are small but not too small. I don't know the exact ideal size but it is what we are told to select when doing the size selection or about 250 bp I think.

It probably also has to do with the way protocols were worked out. The original protocols used the gel purification step to get rid of adapter dimers as well as size select. Now SPRI beads are widely used. I think we will see more and more protocols that eliminate the gel purification step because it is labor intensive, has poor recovery and introduces sample to sample variability.

Suboptimal protocols hang around for a long time in biology simple because they work good enough for what most people do.

[Turnerac0987]

In my old lab we never did the gel size selection step. For our framentation step we used Nextera Enzyme, and before that we used Covaris shearing. I guess those methods gave us small enough fragments to work with where we didn't need to do gel size selection.
I just moved to a new lab and I'm about to start a protocol that wants me to do a gel size selection step and I'm also wondering if it's possible to skip it. I'll be using Covaris to shear so as long as I use the right settings I assume I'll be able to get the right size bands?
I'm also interested in hearing opinions on whether we can skip the gel size selection.

[mboth]

Thanks everybody!
I guess the same applies to the RNA-seq protocol, where having a large variety of molecules is even more important than having them all the same size.
Has anybody ever seen a difference of size-selected and non-size-selected libraries in the actual data?