Impacts of Marker Class Bias Relative to Locus-Specific Variability on Population Inferences in Chinook Salmon: A Comparison of Single-Nucleotide Polymorphisms with Short Tandem Repeats and Allozymes

Abstract

Single-nucleotide polymorphisms (SNPs) exhibit several attributes that make them appealing as a class of genetic markers for applications in ecology and evolution. Two commonly cited limitations of SNPs in this capacity are that ascertainment bias and natural selection may shape allele frequencies of these markers, thus biasing estimates of population structure. The impacts of ascertainment bias and selection on estimates of population parameters have been demonstrated in a few model species, but their impacts relative to locus-specific variability and other potential complications on structure inferences in wild populations are unclear. We examined 22 allozymes, 9 short tandem repeats (STRs), and 41 SNPs in approximately 1,300 Chinook salmon Oncorhynchus tshawytscha representing 16 collections. We used plots of the genetic differentiation index F _ST versus heterozygosity and sequence criteria to identify SNPs that might be under natural selection. We then calculated several measures of population structure based on the three marker sets and a subset of the SNPs from which loci identified as likely targets of natural selection had been removed. Correlation of genetic distances between collections was stronger between allozymes and SNPs than between either of these markers and STRs, suggesting that the influences of marker class bias (e.g., selection and ascertainment bias) were smaller than impacts of locus-specific effects. Divergence estimates between SNP ascertainment populations were not significantly higher when based on SNPs than when based on other markers. Overall divergence (F _ST) was higher for SNPs than for allozymes; however, the choice of F _ST estimator influenced the relative values for STRs and SNPs. Estimates of within-population diversity based on allozymes and STRs correlated better with each other than with estimates based on SNPs; such estimates based on SNPs were relatively low for collections from populations outside the geographic coverage of the SNP ascertainment sample.