We tend to think of evolution as something that occurs over very long timescales – thousands and millions of years. While it is true that big changes require a long time, smaller changes must be occurring on shorter timescales. It is those small changes that accumulate into the bigger changes.

However, it’s difficult to perceive and measure the smaller changes that occur over short periods of time. Those small evolutionary changes can also be lost in the background noise of natural variations in a population.

What we need are techniques for observing populations in greater detail, so we can track evolutionary shifts. A 2016 study demonstrates a technique for doing just that. The researchers are taking advantage of the age of big data, in this case genetic data. There are now large genomic data sets available, and researchers can plow through that data to look for the telltale signs of evolutionary change, even over as little as a generation.

The authors of the current study developed a technique in which they count single base pair mutations near alleles (an allele is one copy of a gene – we have two copies or alleles of each gene, one from each parent). When genes are spliced together during recombination (the process by which alleles from each parent are randomly assembled into the genome of an egg or sperm) DNA near each allele will tend to go along for the ride. So that nearby DNA can be used to track the history of the allele itself.

The idea is that once a mutation results in a new allele in a population, as the new allele spreads throughout the population (if it does) it will accumulate these singleton mutations in nearby DNA. So when you compare the DNA of individuals in a population, there will be more different such mutations near alleles that have been around for a long time, and fewer different singletons near alleles that are more recent.

Therefore, if you have an allele that is widespread in a population, but with very few singleton differences, that allele must have spread recently and rapidly through that population. That is evidence for selective pressure in favor of that allele. Genetic drift, by contrast, would only result in an allele spreading slowly through a population, accumulating more mutations along the way.

It’s a clever analysis, made possible by these large genomic data sets. What they have found so far is that alleles for fair skin and hair spread in the last 2000 years through the UK population. The analysis does not say what the selection pressure was, it’s just a rough measure of the intensity of the selective pressure. In this case the authors speculate that a fair complexion could have been an advantage in an environment with frequent overcast skies and less sun (for more vitamin D creation). Or it could have simply been sexual selection.

They also found that an allele for a nicotine receptor associated with greater difficulty quitting smoking has been decreasing in the last century. Although as smoking decreases overall, the selective pressure against this allele should also decrease.

The team also looked at a suite of alleles – for large hip size, large head circumference, and height. These should go together, as larger hips are needed to give birth to larger headed babies. They found that all these alleles have been increasing together over the last few thousand years.

So the technique seems to able to see changes from a few decades to a few thousand years. As the genomic data sets grow, it will only become easier to track genetic changes in the human population. At its core, that is what evolution is – changes in allele frequency in a population over time (there is debate about whether this is the best definition for evolution, but I will leave that for another day).

One caveat to this technique, which is generic to any science involving such large data sets, is that it is easy to get mislead by random noise and confirmation bias. If you go out of your way to look for specific patterns, you will likely find them. There is also likely to be random patterns in the noise, and it’s easy to underestimate the odds of finding them. Statistical analysis is extremely tricky, and it can be very easy to fool yourself if you’re not careful.

Large genomic data sets, therefore, represent a great potential, but also a great risk for data mining (like any large set of noisy data). For this reason we have to look skeptically at any results, and I tend to put more confidence in them when they are independently replicated and picked over by other scientists. So while I find this new technique to be intriguing, and it seems legit as far as I can tell, I also consider it tentative until it has gone through the meat-grinder of post-publication peer-review and replication.

I haven’t seen any further publications with this technique since this study came out, even by the original team, but it has only been a couple years. We’ll have to give it more time.

I should also note that this is not the only analysis to find evidence of recent evolution, or the first scientific study to document evolution in action. Lenski’s bacteria research remains a great example of evolution in the lab. There are also studies showing that certain alleles for brain development have been spreading through the human population over the last tens of thousands of years.

We appear to have entered the age of big genomic data, and it’s only going to get bigger. This provides an interesting new window onto population genetics and evolution, and we just at the beginning.