Reader Comments

Post a new comment on this article

What could go wrong with this model?

Posted by GregLaden on 06 Mar 2008 at 02:40 GMT

I apologize if I'm a bit vague in my comment: I read this paper some time ago and am only now commenting.

One of the main contributions of this paper is the linking up actual historical data (arcaheology and palaeoecological reconstructions) with the genetic data. This of course has been attempted numerous times in the past, but here we have the best genetic study so far being put together with very current archeology.

However, this means that there are several different contingencies required for the whole model to work. Where and when various population splits, constrictions, expansions, and movements occur shifts if any of the following changes:

-The meaning of pre 14K arcaheological dates in the New World

-The rate estimates for DNA change. There is no really good reason to believe that these are stable. (The estimates)

-The time and place of the ice free corridor.

I applaud the authors for moving ahead with this synthesis, and I'm not trying to be a pedant. I would simply like to ask, if the opportunity arose, of the authors: Which of the key elements in your model would most thoroughly turn the whole model into mush, showing it to be wrong? The DNA timing? The paleontology?

I think this question could be answered two different ways. 1) Which element is most important and if disproven would be devastating, regardless of how confident the authors are in the argument, vs. 2) which elements of the overall synthesis are least certain because of the current nature of the data.

Any one of us could look in on this paper and come up with suggestions, but I think the most insightful and interesting, and introspective view would come from the authors themselves.

Again, fantastic paper, a true step forward (several steps, actually).

RE: What could go wrong with this model?

conniemulligan replied to GregLaden on 18 Mar 2008 at 13:53 GMT

Response to GregLaden:

It is interesting to speculate on which types of data might be most devastating to our model because we envision our model as being able to accommodate many updated, revised, and/or expanded datasets. Indeed, the point of integrating genetic, archaeological, geological, and paleoecological data was to provide a robust model that did not depend on a single dataset. However, it is always interesting to speculate, so we provide the following possible scenarios that might require a radical revision of our model:
- Drastic revision of mitochondrial genome mutation rates: A change in mutation rates would correspondingly change the x-axes of our skyline plots, i.e. a faster mutation rate would increase the dates and a slower mutation rate would decrease the dates. The same changes would occur on the y-axes, with faster mutation rates decreasing Ne estimates, whilst slower ones decrease Ne estimates. (These linear changes are due to the scalar properties of mutation rates on Ne and time estimates in coalescent analyses.) However, the mutation rate estimates for the human mitochondrial genome are among the most robust evolutionary mutation rate estimates. Specifically, the mutation rate estimates by Ingman et al. (Nature, 2000, 408:708-713) were replicated by Atkinson et al. (Mol. Biol. Evol., 2008, 25:468-474) in an independent analysis. These mutation rate estimates were amazingly close, i.e. 1.70 x 10-8 and 1.69 x 10-8 substitutions per site per year, respectively.
- Future identification of a single earlier (pre-14kya) archaeological date for human occupation of the New World: Only a single earlier date might suggest an occupation that did not contribute to the modern Amerind gene pool, i.e. an unsuccessful occupation, or a population that did not experience a population expansion such as would be detected by our analysis. Our model would remain essentially unchanged because we are interested in explaining the history of major migration and expansion events.
- Future identification of multiple earlier (pre-14kya) archaeological dates for human occupation of the New World: If multiple earlier archaeological sites were discovered, and they were geographically dispersed, this would imply an earlier, and significant, migration that is not predicted by our model. Depending on the exact dates, we would have to examine our data to see if our model could accommodate those dates, i.e. our signal of an expansion event ~15,000 years ago could also be consistent with an entry date many thousands of years earlier if the expansion postdated the entry by thousands of years. However, discovery of multiple, geographically dispersed pre-14kya human occupation sites seems unlikely at this stage of North American archaeology.
- New geological data that Beringia was inundated much earlier than the ice-free corridor in North American opened up: These data seem unlikely since the two events are related, i.e. increasing temperatures caused both the ice to melt exposing the ice-free corridor and the sea levels to rise that inundated Beringia. However, if Beringia was inundated prior to the ice-free corridor opening, this fact might suggest that coastal routes into the New World were the preferred means of entry.