Tuesday, July 31, 2012

Dienekes' Estimates Of Y-DNA Clade Ages

Armed with new and improved software, methodologies and data, Dienekes' has produced his own estimates of the ages of various Y-DNA haplogroups.   I summarize Dienekes' date estimates for Y-DNA haplogroups below with somewhat less technical detail and some more descriptive context from the overall phylogeny of the Y-DNA haplogroups (upon which there is a nearly complete consensus as to locations on a branching tree of descent, although there is not a consensus as to the age and the place of origin of the clades). 


His knowledgeable amateur estimates (his day job is in the IT industry in Greece) are generally in line with previous estimates from published journals.  Also, note that many of the gaps in his estimates to date (which he is rapidly filling) are strongly constrained and predicted by estimates for parent clades, daughter clades and sister clades, and that there are estimates in the literature for these clades, just not estimates within the last couple of days from Dienekes as part of a coherent effort.  The most interesting gaps in the estimates are those for D, G, H, L and T which have the weakest analogs and bounds from the estimates that Dienekes has made so far.


I analyze the issue of the calibration of the Y-DNA genetic based date estimates after recapping his results below.  I also note briefly that the one outlier in his estimates is C1-C3 which one would expect for a variety of reasons to be much older. Analysis of what the dates themselves imply about human prehistory and the differences in the conclusions that I draw from the data compared to those drawn by Dienekes will have to wait for another post or posts.  (See also my November 2010 post at Wash Park Prophet addressing similar issues with quite a bit of reference to the literature.)


Y-DNA Clades, Comparison Ages (in years ago), and Description of Clade.

Y-DNA Adam.  159,298 This is the most recent common male ancestor of all modern humans.
*BT 71,188 The clade that breaks off from the mostly Paleo-African A.
**CT 62,439 Shared root of DE and CF from African B. 


DE 62,205 The source of Asian D & African E. A few basal cases found in both areas.
*D Sporadic Asian distribution (e.g. Andaman Islands, Japan, Tibet).  No estimate yet (less than DE).
*E 57,703 The root of this African clade with migrant branches in West Eurasia
**E1b1 43,587 The predominant sub-haplogroup of E in Africa esp. outside E. Africa.
***E1b1b1a1 13,817-19,482A leading West African clade.


CF 47,379 Root of all predominantly non-African clades except D.
*C1-C3 25,022 Mostly Southern Asian.
*F West Eurasian rooted branches are from F as are many Asian ones. No estimate yet (less than CF and more than IJK)
**G Mostly European/Caucasian branch of F.  No estimate yet (less than CF, similar to IJK)
**H Mostly South Asian branch of F.  No estimate yet (less than CF, similar to IJK)


IJK 41,910 Root of IJ and K.  Derives from F.
*IJ 35,589 Root of I and J.
**I 26,885 West Eurasian, possibly predominant in European Paleolithic
***I2a1 19,513 Mostly Central and Eastern European
**J 24,497 West Eurasian. J1 is mostly Semitic (see below for J2).
***J2 23,420 Highland West Asian root probably spread by Indo-Europeans.
* K-M9 36,389 Root of MNOPS and LT as well as other K clades.


LT Derived from K.  No estimate yet (less than K)
*L Generally Indus Valley.  No estimate yet (less than LT).
*T Broad West Eurasian and North and East African distribution.  No estimate yet (less than LT)


MNOPS Derived from K. No estimate yet (less than K, more than NO and P)
*M Southern Asian. No estimate yet (less than K and similar to NO and P)
*NO 32,467 An Asian branch of MNOPS.
**N Northern Asian.  No estimate yet (should be similar to O).
**O 26,145-27,870 Mostly Southern Asian. See [Note 1]
***O3a1-O3a2 18,765 Northern Asian.
*P 33,043 Central and South Asian.  Sister clade to NO. 
**Q North Asian and Americas.  Descendant of P.  No estimate yet (less than P)
**R West Eurasian.  Descendant of P.  No estimate yet (less than P and more than R1)
***R1 23,657 A predominant European haplogroup.
****R1b1a2a1a1a5 6,476 About 25% of (mostly) Western European haplogroups within R1b.
***R2 Generally Indus River Valley.  No estimate yet (similar to R1)
*S Melanesian. No estimate yet (less than K, similar to NO and P)

[Note 1]: "This haplogroup [O] appears in 80-90% of most of populations in East Asia and Southeast Asia, and it is almost exclusive to that region: M175 is almost nonexistent in Western Siberia, Western Asia, Europe, and Africa and is completely absent from the Americas, although certain subclades of Haplogroup O do achieve significant frequencies among some populations of South Asia, Central Asia, and Oceania."  Sub-branches of haplogroup O have strong language family associations.


The Case That His Estimates Need To Be Calibrated To Be 50% Longer

His estimates using the methods he is using, like the estimates in most published academic work, are low compared to the estimates one would make looking at the more definitively dated archaeological record.


A date about 50% higher would be a better fit to the archaeological evidence of modern humans which has been estimated at as much as 250,000 years ago. A 50% higher date also seems to be a better fit for DE and CT which should match the Out of Africa date which archaeologically is closer to 100,000 years ago than 62,000 years ago. This would also put the sister clades of MNOPS closer to the archaeologically supported date of around 45,000 for Melanesia and one would expect the sister clade NO to have an age similar to Melanesian S. A 50% longer date would have I coincide with the modern human settlement of Europe, which would be a more likely date than one six thousand years before the Last Glacial Maximum in the Upper Paleolithic. This would put the most common haplogroup of R1b in Western Europe in the Holocene, fairly close in time to earliest ancient DNA evidence of modern mtDNA haplogroups like H in Europe and North Africa, although the shorter estimate would be a fit to the European Copper Age which would also make some sense.


A recalibration of Y-DNA dates would also bring Y-DNA haplogroup age estimates more closely in line with age estimates of mtDNA haplogroups that very likely came into being and spread with corresponding Y-DNA haplogroups.  For example, a common estimate for the most recent common maternal ancestor of all modern humans is 200,000 years ago: "In 1987, Cann et al. suggested that mitochondrial Eve may have lived between 140-280 thousand years ago." While, mitochondrial Eve and Y-DNA Adam don't have to have been contemporaries, a 40,000 years spread between the two seems unlikely. The mtDNA estimates for characteristic Out of Africa mtDNA haplogroups are also a bit short relative to the archaeological evidence of humans leaving Afica, but by a lesser amount. Increasing conventional mtDNA estimates by about 20% brings them better into line with the archaeological evidence.

In short, I suspect, given the comparable percentage of mismatch of archaeological dates that ought to correspond with Y-DNA genetic based dates, that the entire scale is miscalibrated and should be about 50% longer, give or take (all of these estimates have meaningful margins of error).


Since Y-DNA estimates are derived from number of generations and then converted to years, generally based on an assumption of a 25 year generation, a recalibration of the data could result from something as simple as the historical average generation length being closer to 17 years than 25 years, which honestly makes more sense in an environment in which there would be no meaningful contraception and no capacity to segregate young adult women from men for many years.


If I recall correctly, the 25 year generation length has ethnographic roots in modern hunter-gather generation length, but the modern hunter-gatherers live in marginal areas where there may be a tendency towards longer generations, while Paleolithic hunter-gatherers would have lived in prime territory that would have made shorter generations feasible. This recalibration may not be equally valid for R1b1a2a1a1a5, however, since it, unlike all of the other dates, derives mostly from people living a food producing life, rather than people living a hunting-gathering life.

NOTE: Repeatedly updated for accuracy, clarity, and formatting on July 31, 2012.

4 comments:

terryt said...

"**O 26,145-27,870 Mostly Southern Asian. See [Note 1]"

I think you've made a misprint here as your note says, 'although certain subclades of Haplogroup O do achieve significant frequencies among some populations of South Asia, Central Asia, and Oceania'. Although all three O haplogroups are present in South Asia (I have doubts about O1 though) they are present only as downstream mutations. O2 is almost, if not entirely, O2a1 and O3 entirely confined to, probably, recent Tibeto-Burman speakers. So O is East Asian. Even in SE Asia it is most likely a relatively recent immigant, perhaps not much more than 10 kya.

"A date about 50% higher would be a better fit to the archaeological evidence of modern humans which has been estimated at as much as 250,000 years ago. A 50% higher date also seems to be a better fit for DE and CT which should match the Out of Africa date which archaeologically is closer to 100,000 years ago than 62,000 years ago".

I'm beginning to think that as I look again at his post. If his results arew taken literally they are much more recent than the dates Behar et al gave for mt-DNA spread. We would have women moving into marginal regions unaccompanied by any men.

"This would put the most common haplogroup of R1b in Western Europe in the Holocene, fairly close in time to earliest ancient DNA evidence of modern mtDNA haplogroups like H in Europe and North Africa, although the shorter estimate would be a fit to the European Copper Age which would also make some sense."

The Copper Age is a possibility. But Dienekes seems intent on proving R1 is recent in Europe. I tend to disagree with him on that although in his latest post he may be about to concede it is ancient in Europe.

Andrew Oh-Willeke said...

Re: Y-DNA O - I am quoting wikipedia on that point. I don't disagree that O in South Asia is the most recent and least admixed of the Y-DNA haplogroups in the South Asian mix (probably in the Bronze Age or later) and that it is mostly found in Tibeto-Burman speakers there. I have no clear opinion on when O arrived in SE Asia or for which direction - but the quote refers to the current distribution, not the historical one.

As to R1b in Western Europe, I very much doubt that it is older than about 12,000-16,000 years and it probably was not there pre-LGM, although I could be convinced that it arrived a few thousand years before the Neolithic did.

Andrew Oh-Willeke said...

Note that Dienekes' revised his estimate for C1-C3 was to 35 kya in a subsequent post (apparently due to a coding issue) with a standard deviation of 8 ky, and was based on a very small sample and he suggests that it would probably be older with a more complete sample.

terryt said...

It's possible that the C1-C3 split is only something like 35 kya. though. But that would mean that C1'3 is a basal C haplogroup, contradicted by our present state of knowledge. But an age of 35,000 years fits a postulated ancient human arrival in Japan.

Dienekes has also lengthened his time of split between C and F to 56,000 years. That leaves plenty of time for C to reach Australia by 45-50 kya. It also fits with the Behar suggestion that mt-DNA N coalesced at 58,859.9 years ago (approximately!). However you could be correct in your idea that the ages are under-estimated.