Nicholas Wade is reporting on an upcoming paper by Chad Huff and Lynn Jorde: "Genome Study Provides a Census of Early Humans".
The Utah team based its estimate on the genetic variation present in two complete human genomes, one prepared by the government’s human genome project and the other by J. Craig Venter, the genome sequencing pioneer. The government decoded a single copy of a mosaic genome derived from a medley of people, apparently of European and Asian origin. Dr. Venter decoded both copies of his own genome, the one inherited from his father and the one from his mother.
The Utah team thus had three genomes to work with and looked at ancient elements known as Alu insertions, the youngest class of which appeared in the human genome around a million years ago. The amount of variation seen in the DNA immediately surrounding the Alu insertions gave a measure of the size of human population at that time.
Their estimate agrees almost exactly with an earlier one, also based on Alu insertions but with sparser data. The insertions tag ancient regions of the genome that are unaffected by the recent growth in population, Dr. Huff said.
I'll probably write some more notes on this when I can get a copy.
At the moment I think it's worth pointing out that the lede of Wade's story is exactly backward. The story is all about how the effective size estimate, 18,500 effective people, is very low. But in reality that's a high estimate compared to what most human geneticists have assumed, only 10,000 individuals.
Neither estimate is really news. Observations in the early 1970's established that 10,000 was around the right order of magnitude for human effective population size. Around 10 years ago, some gene systems, including Alu insertions, appeared to support a higher estimate of effective size up around 18,000 individuals. That still seemed pretty small in evolutionary terms, and didn't change anybody's ideas about ancient population bottlenecks.
The differences between these estimates have never really been resolved. As more and more genes got sequenced, human geneticists seem to have just standardized on the small estimate of 10,000 effective individuals -- even as they started to apply more and more complicated computer models to try to derive estimates of expansion and bottleneck times. (I wrote about the problem of effective population size last year, "Cultural impedance, demographic growth, effective population size".)
A few years ago we started to get good effective size estimates for other primates. As Wade's article points out, the genetic variation of chimpanzees and gorillas lead to estimates of effective size on the order of 25,000 or so individuals. Geneticists noted that these species are therefore much more diverse than humans, with our puny effective size of around 10,000 individuals. Only bonobos seem to be close to the low human value.
Well, if Huff and Jorde are right, human variation is a lot like the amount of variation in chimanzees and gorillas. Those other apes have lived in geographically structured subspecies spanning tropical Africa for several hundred thousand years.
Or have they? Maybe there were massive bottlenecks and population replacements among chimpanzee subspecies. Maybe there was a recent "out of Congo" migration that accounts for the low genetic variation of bonobos. Maybe chimps themselves derive recently from some part of their current range.
Or, maybe the human effective population size isn't so probative.
In any event, the genomes here are all Eurasian. I wonder how much African genomes will increase the diversity? Could it be that we're even more diverse than chimpanzees?