Effects of Sample Selection Bias on the Accuracy of Population Structure and Ancestry Inference

Abstract

Population stratification is an important task in genetic analyses. It provides information about the ancestry of individuals and can be an important confounder in genome-wide association studies. Public genotyping projects have made a large number of datasets available for study. However, practical constraints dictate that of a geographical/ethnic population, only a small number of individuals are genotyped. The resulting data are a sample from the entire population. If the distribution of sample sizes is not representative of the populations being sampled, the accuracy of population stratification analyses of the data could be affected. We attempt to understand the effect of biased sampling on the accuracy of population structure analysis and individual ancestry recovery. We examined two commonly used methods for analyses of such datasets, ADMIXTURE and EIGENSOFT, and found that the accuracy of recovery of population structure is affected to a large extent by the sample used for analysis and how representative it is of the underlying populations. Using simulated data and real genotype data from cattle, we show that sample selection bias can affect the results of population structure analyses. We develop a mathematical framework for sample selection bias in models for population structure and also proposed a correction for sample selection bias using auxiliary information about the sample. We demonstrate that such a correction is effective in practice using simulated and real data.

Dating genomic variants and shared ancestry in population-scale sequencing data

“…American ancestry groups are typically younger than 750 generations (15,000 to 22,500 years), consistent with existing knowledge about the settlement of the Americas via the Bering land bridge that connected Asia and North America during the last glacial maximum around 15,000 to 23,000 years ago [45,46]. We note, however, that recent admixture and the sampling strategies of the different data sets [47,48] can have a strong impact on age distributions. For example, variants at high frequency within American populations but that are nevertheless restricted to just American and African populations are, on average, younger than lower-frequency variants (within American populations) with the same geographical restriction (S6 Fig).…”

Dating genomic variants and shared ancestry in population-scale sequencing data

“…Variants restricted to American ancestry groups are typically younger than 750 generations (15,000 to 22,500 years), consistent with existing knowledge about the settlement of the Americas via the Bering land bridge that connected Asia and North America during the last glacial maximum around 15,000 to 23,000 years ago [42,43]. We note, however, that recent admixture and the sampling strategies of the different data sets [44,45] can have a strong impact on age distributions. For example, variants at high frequency within American populations, but which are nevertheless restricted to just American and African populations, are on average younger than lower frequency variants (within American populations) with the same geographical restriction (Supplementary Figure S6).…”

Hybrid dynamics in a species group of swallowtail butterflies

“…7e). However, the limited sampling of these SNP data sets may have biased STRUCTURE results (Shringarpure & Xing, 2014), and we observe very little separation of Foothills region individuals with DAPC (Fig. 3b, c).…”