SNP ascertainment bias in population genetic analyses: Why it is important, and how to correct it

Abstract

Summary Whole genome sequencing and SNP genotyping arrays can paint strikingly different pictures of demographic history and natural selection. This is because genotyping arrays contain biased sets of pre-ascertained SNPs. In this short review, we use comparisons between high-coverage whole genome sequences of African hunter-gatherers and data from genotyping arrays to highlight how SNP ascertainment bias distorts population genetic inferences. Sample sizes and the populations in which SNPs are discovered affect the characteristics of observed variants. We find that SNPs on genotyping arrays tend to be older and present in multiple populations. In addition, genotyping arrays cause allele frequency distributions to be shifted towards intermediate frequency alleles, and estimates of linkage disequilibrium are modified. Since population genetic analyses depend on allele frequencies it is imperative that researchers are aware of the effects of SNP ascertainment bias. With this in mind we describe multiple ways to correct for SNP ascertainment bias.

Multiple genomic signatures of selection in goats and sheep indigenous to a hot arid environment

“…Differences in demographic history (genetic drift, effective population sizes, bottlenecks/expansion, inbreeding and so on) have been implicated but are unlikely to explain our results because demographic history shapes variations across the whole genome, whereas the effects of selection tend to be localized in a genomic region (the selected locus and linked genetic markers) (Akey, 2009). Owing to this, outlier loci tend to be strong candidates for natural selection and they can lead to many false positives/negatives (Lachance and Tishkoff, 2013), particularly when an ascertained subset of SNPs is assayed. Although it would be unusual not to have false-positive SNPs on each array, we discount false positives because we performed an extra layer of more stringent quality filtering following the initial SNP genotype calling.…”

“…Although it would be unusual not to have false-positive SNPs on each array, we discount false positives because we performed an extra layer of more stringent quality filtering following the initial SNP genotype calling. Ascertainment bias, a major drawback of SNP arrays (Lachance and Tishkoff, 2013), can affect selection scans. Among our study breeds, only the Boer goat, Texel and Romney sheep were used in the development of the two chips.…”

High-throughput sequencing reveals inbreeding depression in a natural population

“…High-density genotyping arrays, for example, are usually based on SNPs identified from small "discovery panels" of individuals. This can distort the allele frequency spectrum, as SNPs with intermediate allele frequencies tend to be overrepresented and rare SNPs are often missing (27). As a result, estimates of heterozygosity based on the polymorphic SNPs tend to be inflated, whereas genome-wide heterozygosity will be underestimated because low-frequency SNPs are ignored (28).…”

Bayesian Genetic Association Test when Secondary Phenotypes Are Available Only in the Case Group

“…Then, the case-control ratio in the available sample does not necessarily represent the prevalence of the disease at the population level due to the way the data are collected. There are several attempts to explore or explicitly incorporate ascertainment bias, for example conducting sensitivity analysis or conditional likelihood approach (Lachance and Tishkoff [26] and Haghighi and Hodge [27]). We leave the issue of incorporating ascertainment bias in Bayesian framework as future work.…”