The ability to digest milk sugar lactose as an adult (lactase persistence) is a genetically determined trait that has only evolved recently (on the time-scale of human evolution). Independent mutation events have given rise to at least 4 different alleles that are believed to cause lactase persistence in different human populations. However, the majority of people in the world (approx two thirds) do not carry any of these alleles and so are considered lactase non-persistent. Such individuals often display symptoms typically referred to as lactose intolerance when they consume foods containing appreciable quantities of lactose.
It is highly likely that lactase persistence alleles increased in frequency primarily in dairying populations, as part of a gene-culture co-evolutionary process. In Europe a single allele (-13,910*T) seems to be the main cause of lactase persistence. Understanding when and where this allele originated and how it spread with the culture of dairying was the main aim of the study by Itan and colleagues (PLoS Comput Biol., 2009).
The graphics below show simulations of the spread of dairying and non-dairying farmers, and a lactase persistence-associated allele, in Europe over the last 9,000 years. In these simulations 'X' marks the spot where this mutation. These are only a few of a very large number of simulations that were performed under a range of conditions (concerning the extent of migration / gene flow / natural selection / etc).
Each simulation generates simulated data on the distribution of lactase persistence at different locations throughout Europe, and on the arrival time of farming at the same locations. These simulated data are obviously shaped by the conditions under which the simulations were performed.
We can compare the simulated data to actual 'real world' data on the same things (frequency of the lactase persistence allele at different locations throughout Europe, and on the arrival time of farming at the same locations). Form this we can choose the simulations that generated data that looks most similar to the actual data.
We then apply the following logic: simulation conditions that give rise to simulated data that looks very similar to 'real world' data are more likely to be true than simulation conditions that give rise to simulated data that looks nothing like 'real world' data. The statistical version of this reasoning is called Approximate Bayesian Computation (ABC)
So if we perform many hundreds of thousands of simulations then retain those where the simulated data is very similar to the 'real world' data, we can ask a range of questions about the conditions under which those 'good simulations' were performed. This allowed us to make inferences on a number of interesting factors that shaped the evolution of lactase persistence.
Some of those interesting factors include the time and location of origin of the lactase persistence-dairying gene-culture co-evolutionary process. The map below shows that the best simulations indicate a central European origin around 7,500 years ago. This fits well with the time and location of origin of the Linearbandkeramik or Linear Pottery, which is itself associated with an increase in cattle usage.