Non-Darwinian estimation: my ancestors, my genes' ancestors

Abstract

There is widespread interest in characterizing the organization of human genetic variation around the world from a population perspective. Related to this are attempts to describe the pattern of genetic variation in the human species generally, including "recreational" genomics, the genome-based estimation of the ancestry of individuals. These approaches rest on subtle concepts of variation, time, and ancestry that are perhaps not widely appreciated. They share the idea that there are, or were, discrete panmictic human populations such that every person is either a member of such a population or is an admixed descendant of them. Ancestry fraction estimation is biased by assumptions about past and present human population structure, as when we trace ancestry to hypothetical unmixed ancestral populations, or assign an individual's ancestry to continental populations that are indistinguishable from classical "races." Attempts to identify even individuals' local subpopulations are less precise than most (geneticists included) expect, because that is usually based on a small portion of a person's ancestry, relative to the much larger pool of comparably related ancestors. It is easier to show that two people have some relationship than to show who or where the actual ancestor was. There is an important distinction between individuals' demographic ancestry and the ancestry of their genes. Despite superficial appearances, these interpretations of genetic data are often based on typological rather than Darwinian thinking, raising important issues about the questions that are actually being asked.

Figure 1.

Admixture structure analysis of a worldwide sample. The x-axis represents individuals from populations arrayed geographically roughly from west to east, as labeled above with sample sources identified below. There is a thin vertical bar for each individual, color-coded to represent his/her admixture proportions from “parental” populations. The parentals are shown schematically as circles below the diagram, with arrows indicating a few of their contributions to individuals in the sample (circles and arrows added by us for this paper). These parental representations are not part of the actual sample but are statistically abstracted from it, as if they actually exist (some individuals in some of the populations are statistically assigned 100% ancestry from one of the parental in this particular data set). This analysis was done using the structure-like program Frappe (that employs a different estimation procedure for similar objectives; Tang et al. 2005). Structure analysis figure reprinted with permission from Li et al. 2008, American Association for the Advancement of Science © 2008.

Figure 2.

Illusions can be generated by population-based structure analysis. Individual genotype relationships from the CEPH-diversity 51 global population samples; neighbor-joining similarity trees were constructed from the matrix of pairwise differences (reprinted from Nievergelt et al. 2007). Panel A color codes the major world regions. The analysis correctly grouped individuals on this broad criterion. However, in panel B, the color code identifies the population source (number and text annotation represents different populations), showing intermingling of similarity within geographic regions. For details see the original paper (Nievergelt et al. 2007).

Figure 3.

Gene identity within and between 100 individuals selected from populations throughout the world. Africa (Mbuti, Mandenka), Europe (Orcadians, Bergamo), South Asia (Burusho, Kalash), East Asia (Han, Cambodian), Americans (Pima, Karitiana).