RAD-Seq vs genotyping by sequencing

[Isa]

Hello, I'm new to bioinformatics and nextgen sequencing and was wondering if anybody has any major pros and cons on this techinque. Also, the only provider I found in US for RADseq is Florigenex, does anybody have any references regarding the service they provide?

Thank you!!!

[nucacidhunter]

You can find lots of published work on both methods and variations. Floragenex has a patent on RAD-Seq, so you only will find service offer in places that has licence from them. For more info look at their wehttp://www.floragenex.com/rad-seqbsite:

[Isa]

Thank you nucacidhunter!

[nucacidhunter]

Hi Isa

I will be happy to help if you have any particular questions after reading literature or watching some presentation videos such as: https://www.youtube.com/watch?v=NGvY...X-v1yT&index=3

[SNPsaurus]

Probably the most popular variants are RAD-Seq, ddRAD and GBS. Others in use are 2b-RAD, ezRAD, and nextRAD (bias disclosure, SNPsaurus' method). There are plusses and minuses to each method.

GBS is popular, and particularly useful, when dealing with a population where the genetics are fully known (such as RILs). With GBS, the usual design is to sequence in each sample a few hundred thousand loci out of a million or more possible loci. Because there is little overlap between samples, the full genotype of each sample is imputed by knowing the parental genotype.

ddRAD and GBS has converged slightly over time. ddRAD and GBS were originally different because GBS generated fragments with a single, frequently cutting restriction enzyme and ddRAD used two enzymes, one rare-cutting and one frequent. But now there is two-enzyme GBS, for instance. ddRAD and GBS have fragments that are defined by a cut site on each side. Thus, particularly for ddRAD, accurate size selection is needed to select the number of loci desired and to maintain consistent sequencing of loci between libraries.

RAD-Seq uses a single enzyme to select the loci, and one side of the fragments are defined by a rare cutter and the other by shearing. Thus, size selection is not as important since all loci are present at a range of sizes. The protocol is a little longer than other methods, but my personal theory is that RAD-Seq has gotten a reputation for being difficult because a relatively naive (in molecular biology) evolutionary biologist or ecologist gives it a try and has problems, then switches to ddRAD or other method and it works, but it mostly because they are a little more experienced in pipetting, etc.

nextRAD mimics RAD, but uses a selective primer to choose the loci. Like RAD-Seq, the second side of the fragment is random.

ezRAD is meant to be a way for labs with little equipment to generate a RAD library. After cutting, the fragments are treated so they can be finished off by a facility using a standard kit.

2b-RAD uses type IIb enzymes to generate a short tag around each cut site (these enzymes cut at a distance from the recognition site). It has some interesting properties, such as all loci will be exactly the same size, say 36 bp, so it has a simple protocol without the size selection issues of ddRAD. But, many people are turned off by the short fragment generated, even though most loci will be unique in a genome.

Just to be technical, The University of Oregon has a patent on RAD-Seq with Floragenex being an exclusive license holder.

All these methods sample the genome so the data will always have the potential for bias compared to whole genome sequencing. But most labs are not willing to trade getting perfect data for two samples with WGS instead of sub sampled data for 200 samples. The bias is going to be low if thousands of loci sequenced anyway.

The trade-off with these methods is that they are able to produce thousands of high-quality genotypes in each sample at a low cost without any prior knowledge of the genome, but you cannot design an experiment that specifically targets 1000 SNPs of interest discovered by other means. To do that you need a fixed-content platform like Fluidigm or GoldenGate. Of course, these are expensive to create, and then will miss any alleles not part of the design.

[Niels Wagemaker]

Look at epiGBS : doi: 10.1038/nmeth.3763
epiGBS: reference-free reduced representation bisulfite sequencing
if you are also interested in dna methylation!
It combines GBS with bisulphite conversion of non methylated Cytosines