GBS Adapter Titration Problem - different Enyzms - no Reference Genome

[bioballs]

I am starting to work with GBS following the Elshire 2011 protocol.
At the moment I am doing the adapter titration, therefore measuring the many adapter concentrations on one sample libraries on an agilent bioanalyzer.

1. Can anybody explain that gab in the Adapter Titration supporting Information from Elshire? Why there is a gap between the Adapter Peak (starts ~52sec) and the Library (starts ~56ses) in the “bad example” and why does the library starts directly at ~52sec in the “good example” without adapter peak?

2. For my plant group I don’t have a reference genome. Elshire uses 100ng per individual sample for digestion and has a reference genome. Do I have to use more DNA when I don´t have a reference genome, because I would need more reads? Which is the crucial step that I have to change to get more reads per fragment?

3. At the moment I have Adapter Titration Graphs from the Bioanalyzer that look almost exactly like the “bad example” from Elshire (to much adapter, will decrease concentration further). I used BamH1 instead of ApeK1, so do I expect less fragments? (1) Does the spiky shape of my graph is due to that I have less fragments with BamH1 than I would have with ApeK1? (2) Should I expect my Library with the "right" adapter concentration to start at ~52sec without the adapter peak (now it shows the same pattern like the graphs of Elshire in my question 2)?

4. Concerning my Graph, the fragment length did not change when I increased the in cycle elongation from 30sec to 50sec. Can anybody explain it?


I will be glad if anybody can help me out.

[SNPsaurus]

I can answer some of these questions.
2) If you don't have a reference genome, the big issue is that you don't know how many fragments you will get for a given enzyme combination. The number of reads needed will be the (# of loci) x (desired read depth) / (fraction of useful reads, e.g. not adapter or off-target) x (some extra factor to account for read variation between samples and loci). So you may need more or fewer reads for your genome than the examples in Elshire. To get more reads per fragment you want to reduce the number of fragments. GBS was really designed for low read depth... a similar protocol with more control over fragment # is ddRAD (Peterson 2012).

3) ApeKI has a recognition site of 4.5 nucleotides, so will cut every 300-700 bp typically whereas BamHI has a 6 bp site and will cut every 4 kb. Very few BamHI fragments will be in the size range for cluster formation on a flow cell, and (4) depending on the enzyme you used very few fragments will amplify given that for Taq polymerase it is recommended that extension time is 1 minute/ 1 kb.

I would suggest (if you haven't already done or tried to do) first figuring out if you can what is the likely size of the genome. From that you can make some guesses as to the # of sites, taking into account how much of the genome is methylated repeat stuff that won't be digested. If you want higher read depth with fewer sites, ddRAD and GBS became mostly the same with the introduction of two-enzyme GBS (Poland 2012) so look at that and the Peterson protocol. The downside is careful size selection is essential compared to GBS (or RAD and nextRAD). ddRAD lets you work with a less-frequent cutter like BamHI but keeps the fragment size small by cutting with a 4-cutter as well to define the other end.

[nucacidhunter]

4. Concerning my Graph, the fragment length did not change when I increased the in cycle elongation from 30sec to 50sec. Can anybody explain it?

It could be for two reasons: 1) Lack of large amplifiable fragments (flanked by adapters), 2) efficiency in amplification of smaller fragments during PCR.

You library profile shows reasonable amount of adapter dimers for this method and you have successfully reduced them. The issue is that BamHI is not suitable for your species. Spiky appearance most likely is due to presence of amplified repeat fragments and also could be due to PCR conditions.

[bioballs]

Thank you very much, I hope the Illumina will give me some Reads at all.

In some weeks I will post a adapter titration library graph with ApeK1 digestet sample.

[SNPsaurus]

You mentioned you want higher read depth at each locus. ApeKI could produce several million sequenceable fragments... are you prepared to sequence just a handful of samples per HiSeq lane to get moderate read depth?

[bioballs]

For one of my genus of interest using BamH1 somebody got around 1500-500 fragments of 100-200bp length (CLC Bio Workbench). Average coverage was around 2-5. I actually don´t really know how many reads I finally will need for my phylogenetic softwares and how many reads would be "right" (just started). As I also have everything for ApeK1 and if ApeK1 would produce to many Fragments I can do a double digest as I already have all required Adapter.

First things first, I need to get the library running in a way that I don´t waste money ;-)

[SNPsaurus]

When you say somebody got 1500 fragments, was that from an actual experiment or in silico prediction? If you are trying to make a phylogeny, then you are correct that you don't need many markers.

I urge you to think about the appropriate design of the experiment before trying to get a particular library to look good when that library may not be what you need and may not even be able to get to ever looking good given your genome of interest and the protocol choice.