Effects of marker type and filtering criteria on Q ST-F ST comparisons.

Zitong Li,Ari Löytynoja,Antoine Fraimout,Juha Merilä

doi:10.1098/rsos.190666

Abstract

Comparative studies of quantitative and neutral genetic differentiation (QST-FST tests) provide means to detect adaptive population differentiation. However, QST-FST tests can be overly liberal if the markers used deflate FST below its expectation, or overly conservative if methodological biases lead to inflated FST estimates. We investigated how marker type and filtering criteria for marker selection influence QST-FST comparisons through their effects on FST using simulations and empirical data on over 18 000 in silico genotyped microsatellites and 3.8 million single-locus polymorphism (SNP) loci from four populations of nine-spined sticklebacks (Pungitius pungitius). Empirical and simulated data revealed that FST decreased with increasing marker variability, and was generally higher with SNPs than with microsatellites. The estimated baseline FST levels were also sensitive to filtering criteria for SNPs: both minor alleles and linkage disequilibrium (LD) pruning influenced FST estimation, as did marker ascertainment. However, in the case of stickleback data used here where QST is high, the choice of marker type, their genomic location, ascertainment and filtering made little difference to outcomes of QST-FST tests. Nevertheless, we recommend that QST-FST tests using microsatellites should discard the most variable loci, and those using SNPs should pay attention to marker ascertainment and properly account for LD before filtering SNPs. This may be especially important when level of quantitative trait differentiation is low and levels of neutral differentiation high.

Highlights

Geographical and temporal differentiation in mean values of quantitative traits are of commonplace occurrence in animal and plant populations
They can be sensitive to how rare low-frequency alleles are collected and filtered from the data [30,31,32,33], and how each researcher decides to deal with markers that are in linkage disequilibrium (LD) with each other [34]
The effect of marker variability on FST was clear from the distribution of FST values: microsatellites with few alleles (1–4) were skewed towards extreme FST values and looked most similar to equivalent distributions for single-locus polymorphism (SNP); those with intermediate numbers of alleles (5–8) had the broadest distribution of FST and most strongly differed between the population pairs, whereas the ones with many alleles (9–21) were least diverged across the different population pairs

Summary

Introduction

Geographical and temporal differentiation in mean values of quantitative traits are of commonplace occurrence in animal and plant populations. FST estimates can be affected by ascertainment bias (sensu [17]) when the selected markers are not representative of the variability observed in all sampled individuals They can be sensitive to how rare low-frequency alleles are collected and filtered from the data [30,31,32,33], and how each researcher decides to deal with markers that are in linkage disequilibrium (LD) with each other [34]. To the best of our knowledge, there have not been any studies of these effects of SNPs filtering in the context of QST-FST comparisons, the problem of defining neutral baseline differentiation level has been well recognized in related contexts (e.g. [25,35,36])

Objectives

Methods

Results

Discussion

Conclusion