Friday, July 3, 2009

Natural selection and the human genome -- the last 10,000 years

Many researchers have been interested in combing the human genome for evidence that natural selection shaped modern humans. The idea is that selection would have favored those of our ancestors who were best able to adapt to changing environments over the past 100,000 or even 10,000 years, as they began to wend their way into the entire diversity of the world's environments, most of them very different from our African savanna homeland. How could we not have faced selective environments? And, look at us--at how different peoples native to different continents appear! How else but by natural selection?

And what about the effects of dense, settled populations after the invention of agriculture, including the strong selective force of disease, which could more easily spread with people living close together? Not only that, but in terms of hundreds of thousands of sequence variants across the genome that were chosen for genotyping for their variation rather than function (most probably functionless or without effect on fitness), it should be easy to identify genome regions that show evidence of selection, by virtue of manifesting greater than average difference among world regions.

The availability of large sets of data on variation in the genome, from populations across the globe, as well as methods for doing large-scale analysis of these data has meant a field day for people interested in this question. They have developed and applied a number of methods for the purpose, to distinguish the statistical chaff from the selective grain: when one does many thousands of tests across the genome one is bound to find 'signals' that arise just by the chance aspects of population sampling. A number of reports have come from these studies, some finding that human evolution is speeding up, and others finding evidence of strong selection at a few sites in the genome. Some claim many positive findings, others that the overall amount of such evidence, given the expectation, is rather small.

A paper in the June issue of PLoS Genetics by Jonathan Pritchard at the University of Chicago and collaborators, reports that selection has not been nearly as strong as many people expected. These results are not a total surprise given that most traits are affected by many genes, that environments change rapidly enough that one era's successful adaptation may be another's maladaption, and that a fundamental survival tactic of all organisms is the ability to adapt behaviorally to changing environments as they happen. Humans above all organisms have the added benefit of culture, which can buffer ecological changes. We make fire, and clothes, to protect against the cold, we invent weapons to substitute for weak teeth in hunting prey, and we have language to organize group responses to threats, the need for food and shelter, and so on.

That doesn't remove humans from the effects of selection, and crowded, permanent agricultural settlement might be particularly vulnerable to new forces, such as infectious pathogens.

But even then, and even if selection is particularly strong (and a 1% selective advantage, often literally too small to be detectable from samples at any given time, is considered strong), it may be difficult to detect selection at the gene level. If many genes affect a trait, and given the clear fact that in a population there are many sequence variants at any gene (and the nearby parts of DNA that regulate its cellular usage), the impact of selection is distributed across these many possibly advantageous variants.

What is favored or disfavored are combinations of variants at many genes. Individually, they have rather trivial effect. And most selection is weak, at least in this sense, as we've written before. When the net advantage of a variant is very small, its frequency changes largely by chance (genetic drift). As a result, when regions of the world are compared, there is usually no difference between the pattern of random markers and genes that have been affected by selection.

Occasionally, a variant has a strong effect in its current environments, and can be quickly advanced in frequency (sickle cell as protection against malaria is one of only a handful of truly convincing examples). But such variants are usually very rare, and won't be seen in the kinds of the small population samples that have been used for geographic studies of human variation. A rare variant that is quickly favored will rise to high frequency and then its effect on the favored trait will become close to average. But at least the pattern of variation in the gene and its nearby chromosome region will show less variation within its source population, and greater differences between populations, than is typical of the rest of the genome. That is the kind of signal people hope to find, but that seem for reasons just described, to be rare in humans.

So, even from a strongly darwinian point of view, 'signatures' of selection will be expected to be weak, as the authors of this paper found. Darwin's work was about phenotypes (traits) not genotypes. The focus on genes may be showing that that is the correct viewpoint in which to understand evolution, whether selection is strong or weak, persistent or occasional.

No comments: