Re: Dating with DNA? Problems and Current Approaches

There are many problems with using DNA to date anything.

First is the use of STR variances. The problem here is that
the variance is strongly influenced by population bottlenecks. This effect appears to be up to a factor of four. For the Y chromosome and some autosomal markers the
actual rate is well known from father-son and genealogical studies. To get the "right" answer from these markers at
several thousand years requires assuming a factor of 4
slower rate. But this depends on unknown population dynamics. This applies to anything that looks at diversity.

This problem does not arise with SNPs if you have enough of them that all lines differ by vary many. You can just use the average rate. But what is the rate? It's too slow to
directly measure, and worse, it is highly variable between sites (on the DNA that is, not places on the planet.) Some of these difference ratios can be well calibrated, some can't. But the overall rate is exceedingly poorly known, except possibly for human mtDNA, where there are some real measurements. By known, I don't mean circularly known, assuming a rate just to get the "right answer". I mean testing pedigrees where the number of generations is known exactly. Eventually this first order killer problem will be solved, and then the second order problems will rear their heads.

Eventually we will be able to date Y haplogroups quite accurately (say +- 20%), but this will take many, many near full-genome sequenced Y chromosomes. These dates fall into the very useful 0-100kYr range. Enough mtDNA full genomes
have been sequenced (mine included, for example) that we KNOW for certain that no +-20% quality measurement of their
origin date will ever be attainable.

How well autosomal pseudo-haplogroups (or point mutations in them) will be dateable is still an open question, as it depends on many things like how long haplogroup blocks we can find. It's unguessable at present.

Doug McDonald
.

Quantcast