Important documentation:

RAD-Seq

What is RADSeq data?

RADSeq is a technique for creating a reduced representation of genomic variation of a given set of samples. Restriction endonucleases are used to randomly fragment genomic DNA, which is followed by several different possible fragment selection steps.

The core concept of RADSeq is that we want to sequence a random subset of the total genomic DNA to obtain genome-wide variation at a fraction of the cost of whole-genome sequencing.

What does it look like?

The “Original” RAD protocol looks something like this:

png

Figure from Andrews et al 2016.

Variants of RAD

Many protocols exist that generate fragments in numerous different ways, including using 1 or 2 restriction enzymes, with or without a PCR step, including multiplexed barcodes to further increase sample throughput, and so on. It’s also possible to sequence a library as single-end or paired-end, adding further information, but also somewhat further complications in assembly. Here are a couple more conceptual diagrams for different protocols (again from Andrews).

png png

Thoughts on informativeness

There is a pervasive misconception that “stacks is for population genetic scale” and “ipyrad is for phylogenetic scale”. I don’t know where this idea came from but it’s folklore. ipyrad can assemble data from within population samples to data including deeply diverged lineages (>50my). Please help me dispel this myth if you encounter it in the wild.

Thoughts on missing data

RADSeq is characterized by missing data, yet people still have a tendency to want to treat it as large-scale multi-locus data. Often this manifests as setting the min_samples_locus parameter to remove 80-90% of missing data. This is bad practice in most cases, and results in throwing away most of your data. The ipyrad analysis tools allow you to make the best use of the data by implementing replicated subsampling, for instance.

RADSeq is not multi-locus data.

Why use ipyrad (Eaton & Overcast 2020) at all?

References