The Human Admixture Map is based on a paper published in Science about 2 years ago. When it came out I wrote a summary on another site to explain in simple terms how it works, so I'll re-post it below for my fellow WikiTree members. My answer is a bit long, so I'll break it into 2 separate answers
A Genetic Atlas of Human Admixture History was published on 14 FEB 2014 in the journal Science.1 The article is, unfortunately, not OpenAccess,2 however they have an interactive map free! showing the results from their study along with a very helpful FAQ (PDF) that provides explanations.
The tl;dr / summary of their study is that a statistical tool was developed to determine how long ago two populations mixed genetically, and used this to create "a global map detailing the genetic histories of 95 different populations across the world, spanning the last four millennia" [LiveScience] using genetic data for 1490 individuals. The algorithm that they use is able to determine approximately when the mixing occurred and the proportion of the population involved.
It's also been covered by the media. The New York Times's article eloquently explains the process:
The team led by Dr. Myers has developed a statistical technique for identifying the chromosomal segments with particular precision. This enables them to perform a second feat, that of assigning a date to the one or more mixing events that have affected a population.
The dating system is based on measuring the length of chromosome segments of a particular ancestry that occur in a population. When people of two different populations intermarry, their children’s genomes carry large chunks of DNA of one parent’s ancestry interspersed with large chunks from the other’s.
In each successive generation, the average size of the chunks becomes smaller because when DNA is swapped between the parents’ genomes in making the eggs or sperm, the cuts needed to generate the swapped sections are made in different places. Therefore, from the average size of the chunks in a person’s genome, the geneticists can calculate the number of generations since the mixing event.
What's also worth noting is that although the technique is not calibrated using actual events, the results do correlate very well with several known historical events.
Their ability to date events and match different populations results from the biological process of chromosomal crossover, which occurs during meiosis (i.e. the production of sperm and egg), results in blending of chromosomes, such that part of a chromosome inherited from one parent gets swapped with part of a chromosome from the other parent resulting in hybrid chromosomes in one's eggs/sperm. Here's a beautiful illustration of how this works and how it affects one's DNA. Each generation the chromosomes get more and more mixed up, resulting in a patchwork of DNA segments from different sources. They utilize the distribution of lengths of each of these DNA segments to calculate how long ago the populations mixed.
Here's an example from their FAQ. It's analogous to what is shown here for an individual, except that (1) does not mark the chromosomes according to recent ancestors (parents/grandparents/etc...), but according to a "world-wide group" and (2) segments for more ancient ancestors would tend to be shorter as the process is repeated. Essentially the amount of crossover and the length of crossover segments from each population in the chromosomes can be used to calculate how long ago the mixing process began.
The authors of the science paper use an algorithm that looks at each chromosome and "paints" it according to ancestry:
takes a string of DNA from the beginning of the chromosome of an individual and colors it (i.e. “paints” it) according to the world-wide group of the chromosome in the sample that it most closely matches. Such a match implies recent shared ancestry with that world-wide group. Each of our chromosomes is a mosaic, stitched together from the chromosomes of our ancestors. The CHROMOPAINTER algorithm also detects the point at which the closest match in the samples changes and paints the next string based on the identity of the new closest match. Hence each chromosome is painted using multiple colors. Because human variation is usually widely shared across populations, this painting does not match the underlying ancestry completely, but does give a hint ... [Source: FAQ]
From the distribution of lengths of those coloured segments they create coancestry curves:
Coancestry curves capture information about the lengths of segments inherited from each original admixing source group. They do this by tabulating how rapidly the painting palette changes as genetic distance increases along the chromosome. [...] Theory suggests that these coancestry curves should follow an exponential distribution with rate equal
to number of generations ago that the admixture event occurred. [Source: FAQ]
[Note: coancestry curves are the little boxes shown on the right side of the interactive map]
They demonstrate how this works in a *simulated* run where they digitally simulate a mixing event (30 generations prior); create a mixed chromosome; then have their algorithm reverse-calculate the parameters for the mixing event, for which they *correctly* calculate an event at 30 generations (30 ± 3, 95% confidence) from the marked sources. [Figure].