How and how much does RAD-seq bias genetic diversity estimates?

Area of Science:

• Population genomics
• Molecular evolution
• Bioinformatics

Background:

• RAD sequencing (RAD-seq) is a widely used tool in population genomics.
• Previous studies have indicated potential biases in RAD-seq data, particularly underestimating genetic diversity due to polymorphism in restriction sites.
• This bias arises from the preferential sampling of closely related haplotypes.

Purpose of the Study:

• To theoretically clarify the bias in RAD-seq data caused by restriction site polymorphism.
• To investigate the confounding effects of population structure and selection on this bias.
• To develop and test an approximate Bayesian computation (ABC) based method for correcting RAD-seq bias.

Main Methods:

• Theoretical modeling of RAD-seq bias under neutral and panmictic conditions.
• In silico digestion of full genomes to simulate RAD-seq data.
• Comparison of model predictions with simulated data from Drosophila melanogaster and Shizophyllum commune.
• Application of ABC methods for bias correction.

Main Results:

• The study confirms that RAD-seq data underestimates genetic diversity, with the bias intensifying at higher polymorphism levels.
• Selection exacerbates the underestimation bias, while spatial genetic structure can mitigate it.
• In silico validation showed that the neutral model partially explains the bias in highly polymorphic species, but with inaccuracies.
• ABC corrections improved estimations but retained some imprecision.

Conclusions:

• The underestimation bias in RAD-seq is more significant in highly polymorphic populations.
• For populations with low polymorphism (e.g., <2%), the bias may be less critical than other sources of error.
• ABC methods, particularly the neutral panmictic model, offer a practical approach to correct RAD-seq bias, though further refinements may be needed to account for complex factors like population structure and selection.