Using Restriction Enzymes pre library prep for targeted sequencing.

[Hoban]

Hey Everyone,

I'm using a targeted capture approach for sequencing the variable regions in Ig, we're trying to develop a basis for genotyping experiments for Ig. We're going with HiSeq, we have too many samples for PacBio. Ig is hard because of the huge amounts of structural variations that can occur (frequent insertions, deletions, duplications, and 'complex' events) which make sequencing with NGS difficult.

Here's my plan, design an enzyme cocktail that will chop at defined regions in the Ig region from one sample into ~1kb+ segments with the majority of segments in the 1kb-20kb range. Size select (I'm proposing <1kb, 1kb-5kb, 5kb-15kb, 15kb+) and separately fragment -> index each size pool. Then pool everything together (and include a non-restricted regular library prepped genome), and do the capture, amp, and sequencing.

The thought is that the extra information (if the read came from a 1kb, 5kb, 15kb, or 15kb+ region) will help differentiate reads during alignment, indel, and read depth analysis. An ex: Reads 1-2-3-4-5 all align to the same area, reads 1-2-5 are 15kb indexed, 3-4 are 1kb indexed, so it is likely 1-2-5 and 3-4 are from separate areas of the region.

I haven't been able to find anything too similar. I've mostly been going off of RADseq papers but since I still want to sequence the whole region I'm just size selecting and indexing separately then pooling, nothing gets thrown out from the region (assuming everything in the region is capture-able by our custom capture). I've considered doing PacBio for 5-10% of the samples to do de novo assembly and align the HiSeq data to that reference (hg19 is a very poor reference for variable regions in Ig).

Any thoughts/advice?

[SNPsaurus]

That's an interesting idea, although I'm not convinced it would get you much past "so it is likely 1-2-5 and 3-4 are from separate areas of the region" although maybe that is all you need?

I'm not fully understanding the costs of PacBio sequencing of small regions. My intuition is that capture of long DNA fragments and PacBio sequencing will be cheaper and more directly return applicable data compared to individually digesting, size selecting, and fragmenting with indexing of each sample followed by capture. I guess multiplexing the capture is a big help, though. On the other hand, the cost of analysis would be much higher with your protocol and require lots of manual annotation.

There were some genome assembly papers that partitioned the genome into different size restriction fragments to limit the reads for de novo assembly. I can't remember them exactly, but they might have some useful perspective on how much such information can help if you can search them out. One more recent one is http://genome.cshlp.org/content/20/2/249.full but I think some other papers were in this vein earlier on.