Library prep for ATACseq

[viverum]

Hi everyone,

apologies if this has already been asked.

So I recently started reading into how to prepare libraries for ATAC-Seq and I kind of got conflicting information on what the shape of the library is supposed to look like.
From what I thought to know, I'm supposed to see peaks in the fragment length distribution at roughly n*200bp which correspond to enriched cutting sites next to mono, di, tri, ... nucleosomes, refer to https://www.ncbi.nlm.nih.gov/pubmed/24097267, Figure 2. These can be seen in Tapestation/Bioanalyzer traces as a QC and can be recovered from the mapping of paired-end reads.
My question is, as far as I know Illumina (or at least the people in our core facility) recommend to only sequence libraries of 200-500bp insert size on their machines, as longer fragments do not efficiently produce clusters in the flow cell. But does this mean that you lose most of your longer fragments during clustering? Is there a bias toward the very small ones? Or is it a matter of the sequencing protocol and machine that are used? I also heard that some people optimize their library prep to mainly produce short homogeneous libraries which to me sounds like over-digesting the chromatin.

If someone can shed a bit of light on these issues I would be eternally grateful

Cheers

[nucacidhunter]

In Illumina systems fragments up to 1Kb can be clustered and sequenced. But there is a bias toward sequencing smaller fragments. To sequence more large fragments in a library with wide size distribution, sequencing depth need to be increased which might increase duplicate read numbers of shorter fragments.

[pmiguel]

As nucacidhunter stated, there is a bias towards shorter library fragments during clustering.

Also, keep in mind, that the bioanalyzer and (I presume) tapestation signal scale with number of bases, not total molecules. So one 1 kb molecule will produce the signal of two 0.5 kb molecules.

6 years ago I looked into this in some depth with some fairly long-insert PE libraries run on a MiSeq of ours. The basic story is here:

http://seqanswers.com/forums/showpost.php?p=76312&postcount=4

Or, for those unwilling to click on links, when I transformed bioanalyzer data to subtract adapter lengths and correct from mass to molarity I got this graph of the library:



x-axis is insert size in bases and y-axis is arbitrary, but presuming I did my math right, should be linear with numbers of molecules at the specified size.

Whereas, after clustering the library above on an Illumina instrument and mapping the reads back to the genome from whence the library derived, I got this:



x-axis is insert size in bases and y-axis is cluster count at that size.

That shows the degree of bias towards shorter amplicons in something approaching a best-case scenario for an Illumina instrument -- low cluster density on a MiSeq.

--
Phillip

[viverum]

Thanks a lot for your help!