PDA

View Full Version : Iain McDonald's New P312 Analysis



Webb
07-24-2017, 03:00 PM
Dr. McDonald sent out a group email this morning that he has completed his initial phase of dating P312 and clades, applying the same method he used for U106.

rms2
07-24-2017, 03:20 PM
Here (http://www.jb.man.ac.uk/~mcdonald/genetics/p312/tree.html) is the link to his P312 Tree.

MitchellSince1893
07-24-2017, 03:29 PM
Link to all of McDonald's human genetic related work (U106 and P312)
http://www.jb.man.ac.uk/~mcdonald/genetics.html

And his newly released
List of the ages derived for R1b-P312 clades, including their 95% confidence intervals.
http://www.jb.man.ac.uk/~mcdonald/genetics/p312/table.html

George Chandler
07-24-2017, 04:06 PM
It will be interesting to see what future carbon dating tests reveal.

My estimated founder dates still haven't moved much:

DF13= 4,500 years +/- 200yrs (founder location theory modern Belgium)
L21= 4,900 years +/- 200 yrs (founder location theory modern Rheine region Germany)
P312= 5,500 years +/- 300 yrs (possibly modern Romania)

George

rms2
07-24-2017, 04:30 PM
I think you should move your P312 likely point of origin east onto the Pontic steppe.

5,500 years +/- 300 yrs predates Yamnaya in Romania. You'd be getting into Gimbutas' Kurgan I and II periods.

Dewsloth
07-24-2017, 04:40 PM
This (from his site) looks funny:


The table below gives a "best guess" at a convergence date, but the true date could be anywhere within the stated 95% confidence interval (and even then only with 95% certainty).
Clade\t Best guess 95% confidence interval
P312 3155 BC (3898 BC — 2568 BC)
DF19 2747 BC (3596 BC — 1773 BC)
DF87 1159 AD (432 AD — 1602 AD)
BY19316 1589 AD (1136 AD — 1857 AD)

The first split under DF19 is between Z302/DF87 and DF88. Usually Z302 and DF87 are grouped together, but DF88 is clearly FAR older (even 6DRIF23 from ~200AD is many subclades below DF88).
So does that imply that while Z302 may be about as old as DF88, that every known line of Z302 also has a common ancestor (DF87) from ~1100AD?

George Chandler
07-24-2017, 04:50 PM
I think you should move your P312 likely point of origin east onto the Pontic steppe.

5,500 years +/- 300 yrs predates Yamnaya in Romania. You'd be getting into Gimbutas' Kurgan I and II periods.

I'm thinking right dates... late Western Black Sea late Cernavoda culture. Only my personal theory with the evidence I've seen for P312 with a fast move west with the majority of P312 taken the more southern route and the majority of U106 taking the more northern route.

Webb
07-24-2017, 05:24 PM
These date estimates makes DF27 the grand papa of P312, which is interesting, and not at all surprising.

kinman
07-24-2017, 05:25 PM
The "best guess" age for U152 seems to be far too young compared to its brother clade DF27 (both of which share ZZ11). Surely they didn't originate 300 years apart. Furthermore, that same "best guess" age for U152 also places it extremely close to the age of its subclade L2.


Link to all of McDonald's human genetic related work (U106 and P312)
http://www.jb.man.ac.uk/~mcdonald/genetics.html

And his newly released
http://www.jb.man.ac.uk/~mcdonald/genetics/p312/table.html

MitchellSince1893
07-24-2017, 06:39 PM
The "best guess" age for U152 seems to be far too young compared to its brother clade DF27 (both of which share ZZ11). Surely they didn't originate 300 years apart. Furthermore, that same "best guess" age for U152 also places it extremely close to the age of its subclade L2.
I brought this very issue up to Dr. McDonald last month.

His response

I will also draw attention to the uncertainties, which are an integral part of the analysis, but unfortunately one of the hardest parts to communicate accurately to a non-scientific audience. The two entries for U152 and DF27 are:
Clade Best guess 95% confidence interval
U152 2737 BC (3320 BC — 2233 BC)
DF27 3028 BC (3741 BC — 2423 BC)
So the "best guess" lies within a possible range that spans many centuries. While ages do correlate with each other within the range of uncertainty, the accuracy of the BigY tests generally isn't sufficient to make the distinction between which of these came first. It takes a more extensive test, probably along the lines of the Oxford Nanopore tests that are still under development.

McDonald emphasized that one should pay more attention to the date range rather than best guess date based on the uncertainty involved.

The take away for me is that the overlap of date ranges that DF27 and U152 share i.e. 3320 BC to 2423 BC (we can move 2423 BC back to no later than 2512 BC due to the known RISE563 date).

This tells me that DF27 and U152 probably formed in the 3320 BC to ~2700 BC range...2700 instead of 2512 BC because I doubt we hit the lottery and RISE563 was the original man with U152 SNP mutation.

rms2
07-24-2017, 06:59 PM
I'm thinking right dates... late Western Black Sea late Cernavoda culture. Only my personal theory with the evidence I've seen for P312 with a fast move west with the majority of P312 taken the more southern route and the majority of U106 taking the more northern route.

I would hesitate to be that specific. A P312 origin on the Pontic steppe leaves P312 free to have gone into Corded Ware via the North European Plain or to have gone around the south end of Carpathians, up the Danube, and into the Carpathian Basin.

IMHO, Romania is too far west at that early date.

kinman
07-24-2017, 07:06 PM
I personally think the DF27 confidence interval is probably more accurate, so I would widen that to 3741 BC to 2700 BC.

But either way, it certainly calls into question YFull's current best guess of 4400 years before present (confidence interval 4800-3900 ybp; which is roughly 2800-1900 BC). Both 3320 and 3741 BC would be way out of their 95% confidence interval (by 5 or 9 centuries). I don't trust their numbers at all.


I brought this very issue up to Dr. McDonald last month.

His response


McDonald emphasized that one should pay more attention to the date range rather than best guess date based on the uncertainty involved.

The take away for me is that the overlap of date ranges that DF27 and U152 share i.e. 3320 BC to 2423 BC (we can move 2423 BC back to no later than 2512 BC due to the known RISE563 date).

This tells me that DF27 and U152 probably formed in the 3320 BC to ~2700 BC range.

George Chandler
07-24-2017, 08:37 PM
I would hesitate to be that specific. A P312 origin on the Pontic steppe leaves P312 free to have gone into Corded Ware via the North European Plain or to have gone around the south end of Carpathians, up the Danube, and into the Carpathian Basin.

IMHO, Romania is too far west at that early date.

I know the theory is a bit more controversial given some pottery details (even though they do have some characteristics of steppe pottery). The use of hill fort settlements is interesting in this small geographic area. Right now with everything I've seen from others (personal and private) I would be looking in that area. You could be right and it was born farther east but for now I'm thinking they were part of the mass movement and whatever pushed them out the P312's that moved north with the corded were the ancestors of L21 and it would surprise me that the ancestor of U152's took that southern route (the U152 experts may dispute that) through the Balkans, Northern Italy etc. It would make sense if a culture were on the move quickly they will possibly adopt cultural traditions of other groups if the interactions are peaceful.

Like I said I know it isn't a theory without argument and I accept that.

George

MitchellSince1893
07-25-2017, 01:15 AM
Message from Dr. Iain McDonald

Dear all in P312,
[cc. P312, U152, L21, DF27 and U106 project Yahoo forums - please feel free to link to this post on other websites]

Many of you will already know me for my work in U106. By way of introduction, I am Dr. Iain McDonald, I'm a co-admin of the U106 project. Professionally, I am an astrophysicist, but I have a side interest in applying physical techniques to genetic genealogy, and I currently hold an Honorary Research Fellowship at the University of Strathclyde.

Within U106, I have been working on a set of software which formulates a haplotree and calculates ages for each clade, based on BigY data. Over the last few months, I've been working with Jef Treece to extend this to P312. You can find the results on my website:
http://www.jb.man.ac.uk/~mcdonald/genetics.html
There are several different versions of the reports, depending on how deeply you want to look into the data and in what format you prefer it. The reports are scale-able, so you can zoom in and out of the data on your web browser. The format closely mirrors that of Alex's Big Tree - Alex and I have discussed integrating this analysis into Alex's tree, but we're a long way off being able to do that properly right now.

The idea of this effort is that we can perform an age analysis like YFull's, except with more complete data (because we do it for free). We also use a more refined mathematical approach, but the basis of the calculation remains the same. We are not performing a BAM file analysis, for which I would continue to recommend both YFull or Full Genomes Corp. Consequently, while we should generally do better than YFull in terms of age estimation because of our larger database and better mathematics, if the same data is uploaded to YFull it should normally produce better results for the smallest clades.

Before you use these data, you should be aware of what this analysis does and does not include.

1. This analysis is still in its early stages. There are a variety of effects that this analysis does not correctly take into account, which we are hoping to address these in a new version, but that will take some time to code up. These dates should be mostly accurate (if necessarily imprecise). However, they are subject to change, and you should take them with a pinch of salt while we see whether there are any remaining issues. We hope to merge the U106 and P312 analyses in future, and better incorporate historical information and ancient DNA results.

2. This analysis is not a complete nor definitive haplotree. Alex's Big Tree, or any trees maintained by people within your project, should be more complete and accurate than mine.

3. This is because this analysis only uses BigY tests. With a few hand-coded exceptions, this tree does not contain SNPs or clades that are not well covered by BigY.

4. Only tests uploaded to groups' Yahoo forums before the 6th June have made it into the version currently online. Once we have had initial feedback from this first version, we will work to incorporate the last two months' worth of tests. If you want to be included in this analysis, please download your raw data and upload it to the relevant Yahoo forum using the instructions at the link above.

5. SNPs cannot be ordered within in a clade, unless someone comes along who is positive for some but negative for others.

6. Hence, this analysis does not calculate ages for SNPs. It calculates ages for clades, i.e. the TMRCA (time to most-recent common ancestor) for the BigY testers who share a particular haplogroup. The listed dates are equivalent to the TMRCA dates in YFull.

7. Clades will only show up with an age when more than one test exists within that clade (since, for clades with only one test, the TMRCA = the birth date of the tester).

8. Pay attention to the uncertainty range on the dates. Most clades can only be dated to within about a millennium. Use the "best guess" dates as a guide, but you should quote the uncertainty ranges in any correspondence. About 5% of dates will not be contained within these uncertainty ranges: we currently know that the Royal Stewart line is among these 5%.

9. Some of the reports are too big for many popular web browsers or spreadsheet software packages to open. The reports have been broken up to allow people with other software to see their relevant sections.

10. The most common mistake when opening the full report in Excel (etc.) is to forget to set commas as the only separator when the dialogue box appears.

11. A wider list of FAQs is available at the above link.

Questions?

This report has been passed around the administrators of the main P312 projects, but there will be specific things they have not picked up on, or which I have not explained properly. I welcome any input people have.

I will be monitoring forum replies to these messages until Thursday evening, when I leave for the weekend. After that, you are best to get in touch privately, either via a direct reply to this message, or via the e-mail address listed at the U106 project, or via its Yahoo forum. Before you do, however, please read the FAQ on the website to see if there is an answer to your question already posted. It's possible that I will get a large number of replies to this message, so it may take me some time to go through them. My replies may have to be briefer than I would like, for which I apologise in advance.

Best wishes,

Iain.
__._,_.___

kinman
07-25-2017, 02:49 AM
As the possible date of origin of P312 creeps back towards 3300 BC or even 3500 BC, another thing to keep in mind is that the plague bacterium (Yersinia pestis) is now thought to have evolved about 4000 B.C. in Kazakhstan (not all that far from the Volga River and eastern Ukraine). Therefore, the L23 and L51 ancestors of P312 who had survived the pneumonic plague could have passed along immunological resistance to plague to their P312 descendants.

Those ancestors (along with Z2103) perhaps not only had horses to their advantage in expanding westward, but as carriers of the plague (to which they had acquired some resistance), they could have also decimated sizeable populations of the non-R1b inhabitants of western Ukraine, Moldova, and even Romania by disease (rather than fighting). R1a could have done the same in northern Europe.

Just as the much later Spanish Conquistadors in comparatively small numbers would conquer the Aztecs after giving them diseases like small pox (to which they had no immunity), the Kurgan fighters could have accidentally done the same thing with the plague. And who knows, like the Aztecs, perhaps the conquered local populations even initially regarded the Kurgan newcomers (on their horses) as gods. Disease as an accidental weapon decimated a large number of Native Americans in North America as well.


Link to all of McDonald's human genetic related work (U106 and P312)
http://www.jb.man.ac.uk/~mcdonald/genetics.html

And his newly released
http://www.jb.man.ac.uk/~mcdonald/genetics/p312/table.html

Osiris
07-25-2017, 09:05 PM
Great stuff, thank you very much Iain. It's too bad about some of the scientific literature not being able to be included. Hopefully FTDNA takes off in German and Italy in the near future.

kinman
07-26-2017, 01:01 AM
As the possible date of origin of P312 creeps back towards 3300 BC or even 3500 BC, another thing to keep in mind is that the plague bacterium (Yersinia pestis) is now thought to have evolved about 4000 B.C. in Kazakhstan (not all that far from the Volga River and eastern Ukraine). Therefore, the L23 and L51 ancestors of P312 who had survived the pneumonic plague could have passed along immunological resistance to plague to their P312 descendants.

Those ancestors (along with Z2103) perhaps not only had horses to their advantage in expanding westward, but as carriers of the plague (to which they had acquired some resistance), they could have also decimated sizeable populations of the non-R1b inhabitants of western Ukraine, Moldova, and even Romania by disease (rather than fighting). R1a could have done the same in northern Europe.

Just as the much later Spanish Conquistadors in comparatively small numbers would conquer the Aztecs after giving them diseases like small pox (to which they had no immunity), the Kurgan fighters could have accidentally done the same thing with the plague. And who knows, like the Aztecs, perhaps the conquered local populations even initially regarded the Kurgan newcomers (on their horses) as gods. Disease as an accidental weapon decimated a large number of Native Americans in North America as well.

This early form of plague bacterium did not yet have the gene to allow them to live in fleas, so it was instead passed from person to person (like influenza).
Here is a quote from a 2015 article on this research (and below that is a weblink to the whole article):

"DNA samples revealed that a group of nomadic herders, the Yamnaya, swept into Europe from the plains of today’s Russia and Ukraine sometime between 5000 and 4800 years ago, bringing their culture and, perhaps, the Proto-Indo-European language with them. But archaeologist Kristian Kristiansen of the University of Gothenberg in Sweden wondered whether they also brought disease—and suggested that researchers test the DNA of Bronze Age humans in Europe and Asia to find out."

Source: http://www.sciencemag.org/news/2015/10/bronze-age-plague-wasnt-spread-fleas

Webb
07-26-2017, 05:19 PM
I have noted some interesting observations of L165 under DF27 in Dr. McDonalds time estimates. I will post over in the L165 thread.

rncambron
07-26-2017, 07:11 PM
I have commented on the Yahoo L21 site that the McDonald results do not capture the latest thinking on the P312/L21 tree structure.His dataset is limited hence this latest demarcation of phylogenetic blocks has not come through in his results.
According to Alex W's Big tree there is a 4SNP Phylogenetic block between P312 and the L21 Block.Given that at least one of those 4 SNPs is in a McDonald BED region and can be included in age calculations, L21 must be at least several hundred years younger than the dates quoted in the report.

kinman
07-26-2017, 09:21 PM
I was nitpicking at one aspect of the tree too, but I guess we shouldn't look a gift horse in the mouth. It's really quite good for a first version, and I look forward to newer versions in the future. His dates look a lot better than those on YFull. Thank you Iain for all your hard work.

P.S. Rather than L21 being several hundred years younger, I would guess P312 is a few hundred years older. That's the trouble with YFull's best guess estimates (many of them are too young).


I have commented on the Yahoo L21 site that the McDonald results do not capture the latest thinking on the P312/L21 tree structure.His dataset is limited hence this latest demarcation of phylogenetic blocks has not come through in his results.
According to Alex W's Big tree there is a 4SNP Phylogenetic block between P312 and the L21 Block.Given that at least one of those 4 SNPs is in a McDonald BED region and can be included in age calculations, L21 must be at least several hundred years younger than the dates quoted in the report.

rncambron
07-26-2017, 09:40 PM
You've missed the point of my post.
The allocation of SNPs to Phylogenetic Blocks below the parent does not affect the age of the parent.It affects the age of the children.
If you have an agenda re the age of P312 please don't abuse my observation on the age of L21.

kinman
07-26-2017, 10:05 PM
Oh, I understood the point of your post. I earlier pointed out (following ytree.net) that ZZ11 should be inserted between P312 and U152 (and Z40481 should as well). However, with increased sampling I predict that the estimated age of P312 will be several hundred years older. I wasn't trying to abuse your observation on the age of L21. I just think most age estimates will increase in future calculations.


You've missed the point of my post.
The allocation of SNPs to Phylogenetic Blocks below the parent does not affect the age of the parent.It affects the age of the children.
If you have an agenda re the age of P312 please don't abuse my observation on the age of L21.

Wing Genealogist
07-26-2017, 10:19 PM
....The allocation of SNPs to Phylogenetic Blocks below the parent does not affect the age of the parent.It affects the age of the children.....

I have noted (with his work on the U106 clade, which has been going on for several months) every new addition has a "ripple" effect where it affects the ages of multiple clades both "upstream" and "downstream" of the new discovery. However, it is my understanding the breakdown of a Phylogenetic Block would have a larger affect on the age of the parent clade than the ages of the subclades. To clarify this issue, I have posed this question to Dr. McDonald (as he is not a member of this forum). I will try to remember to post his reply here.

rms2
07-30-2017, 03:02 PM
YFull's new tree, version 5.05, is up: R1b Tree (https://www.yfull.com/tree/R1b/).

It does not look like the age estimates have changed. I really like YFull, but that is somewhat disappointing.

MitchellSince1893
07-30-2017, 03:09 PM
YFull's new tree, version 5.05, is up: R1b Tree (https://www.yfull.com/tree/R1b/).

It does not look like the age estimates have changed. I really like YFull, but that is somewhat disappointing.

Unless they modify their methodology it's not going to change much. I would recommend they have a minimum number of samples in a branch/subclade before they use it for dating. Otherwise you have a branch with 2 members having as much weight as one with several hundred.

kinman
07-30-2017, 04:12 PM
My own lineage shows how ridiculous their age estimates are. Subclades within subclades within subclades within subclades, etc., all having the same 4400 ybp formation and 4400 ybp MRCA:
U152
L2
Z49
Z142
FGC22963
FGC22940


YFull's new tree, version 5.05, is up: R1b Tree (https://www.yfull.com/tree/R1b/).

It does not look like the age estimates have changed. I really like YFull, but that is somewhat disappointing.

rms2
07-30-2017, 04:19 PM
My own lineage shows how ridiculous their age estimates are. Subclades within subclades within subclades within subclades, etc., all having the same 4400 ybp formation and 4400 ybp MRCA:
U152
L2
Z49
Z142
FGC22963
FGC22940

I think characterizing their estimates as "ridiculous" is a little extreme. YFull can only use the data they receive from customers with NGS test results.

I am no expert at SNP-counting age estimates myself. I look forward to ancient y-dna results that give us minimums to work with (i.e., at a minimum, this SNP can be no younger than x).

Of course, the problem with that is that radiocarbon dating is not exactly pinpoint in accuracy.

kinman
07-30-2017, 04:36 PM
Well, maybe if more people ridiculed their numbers, perhaps they would do what Mitchell suggested: modify their methodology.

For what it's worth, instead of 4400 ybp over and over and over and over again, my estimates for age of formation are as follows (maybe a little too early for some to swallow, but at least they make sense with subclades being younger than their parents):

U152 --- 5100 ybp
L2 --- 4900 ybp
Z49 --- 4700 ybp
Z142 --- 4600 ybp
FGC22963 --- 4400 ybp
FGC22940 --- 4200 ybp


I think characterizing their estimates as "ridiculous" is a little extreme. YFull can only use the data they receive from customers with NGS test results.

I am no expert at SNP-counting age estimates myself. I look forward to ancient y-dna results that give us minimums to work with (i.e., at a minimum, this SNP can be no younger than x).

Of course, the problem with that is that radiocarbon dating is not exactly pinpoint in accuracy.

rms2
07-30-2017, 05:03 PM
Well, maybe if more people ridiculed their numbers, perhaps they would do what Mitchell suggested: modify their methodology.

I don't think ridicule is a very effective tool. Better to let them know where and why you think they are wrong and how they could improve.



For what it's worth, instead of 4400 ybp over and over and over and over again, my estimates for age of formation are as follows (maybe a little too early for some to swallow, but at least they make sense with subclades being younger than their parents):

U152 --- 5100 ybp
L2 --- 4900 ybp
Z49 --- 4700 ybp
Z142 --- 4600 ybp
FGC22963 --- 4400 ybp
FGC22940 --- 4200 ybp

You'll have to pardon me, kinman, but whenever someone here writes something like "my estimates", I always shake my head and wonder, "Who are you?".

Why should your estimates be any more reliable than YFull's?

At least some of the guys at YFull are actually geneticists.

Joe B
07-30-2017, 05:38 PM
I couldn't agree with rms2 anymore. The YFull crew are very much out in the open and straightforward with their methodology. It's published for everyone to see. https://www.researchgate.net/publication/273773255_Defining_a_New_Rate_Constant_for_Y-Chromosome_SNPs_based_on_Full_Sequencing_Data
That doesn't mean there is no need for improvement.
Oleg Balanovsky has a nice summation of where we're at now.
Toward a consensus on SNP and STR mutation rates on the human Y-chromosome
https://www.researchgate.net/publication/316553628_Toward_a_consensus_on_SNP_and_STR_mutati on_rates_on_the_human_Y-chromosome

MikeWhalen
07-30-2017, 05:38 PM
Using ridicule to try to make your point with adults is pretty much guaranteed to get you shut down and utterly ignored

Jm2c

Mike

Dewsloth
07-30-2017, 07:27 PM
This (from his site) looks funny:


The table below gives a "best guess" at a convergence date, but the true date could be anywhere within the stated 95% confidence interval (and even then only with 95% certainty).
Clade\t Best guess 95% confidence interval
P312 3155 BC (3898 BC — 2568 BC)
DF19 2747 BC (3596 BC — 1773 BC)
DF87 1159 AD (432 AD — 1602 AD)
BY19316 1589 AD (1136 AD — 1857 AD)

The first split under DF19 is between Z302/DF87 and DF88. Usually Z302 and DF87 are grouped together [snip]


YFull's new tree, version 5.05, is up: R1b Tree (https://www.yfull.com/tree/R1b/).

It does not look like the age estimates have changed. I really like YFull, but that is somewhat disappointing.

YFull:


R-DF19Z4161 * DF19/S232 * S1354/Z8192formed 4400 ybp, TMRCA 4400 ybpinfo
R-DF19*
R-Z39400Z39400 * Z39409 * Z39403+27 SNPsformed 4400 ybp, TMRCA 150 ybpinfo
id:YF08157
id:YF07367ENG
R-Z302CTS12966/DF87 * CTS9798 * Z302/S233formed 4400 ybp, TMRCA 4400 ybpinfo
R-Z302*
id:NA12889CEU
id:HG00138GBR
R-Y23278Z35680 * Z35684 * Z35690+12 SNPsformed 4400 ybp, TMRCA 2500 ybpinfo
id:YF06307
id:YF05568DEU [DE-HE]
R-DF88S4274 * DF88/S4298 * FGC11834/Y3095+5 SNPsformed 4400 ybp, TMRCA 4400 ybpinfo
R-DF88*
id:YF08728NOR [NO-11]
id:YF02816SCT
[snip]
R-S4268Z17111/Y6236 * S4268formed 3900 ybp, TMRCA 3900 ybpinfo
R-S4268*
[snip]
R-Y6237Z17112/Y6237formed 3900 ybp, TMRCA 3200 ybpinfo


Well, I'm still confused. :)

kinman
07-30-2017, 08:26 PM
I was taking courses and doing research in genetics and systematic biology way back in the 1970s, so I do know something about it. I definitely think something is very wrong with their methodology if they continue to have six successive subclades with the same "formed" date:
U152 --- 4400 ybp
L2 --- 4400 ybp
Z49 --- 4400 ybp
Z142 --- 4400 ybp
FGC22963 --- 4400 ybp
FGC22940 --- 4400 ybp

So in this case, I think their estimates are clearly way off. You can't have zero years per SNP between a subclade and its parent clade, much less 6 successive subclades. I agree that you don't need to ridicule them personally, but the above numbers seem ridiculous to me. I am ridiculing numbers, not people. Anyway, I'm sticking with my estimates (which clearly assume a lot more than zero years per SNP).


I don't think ridicule is a very effective tool. Better to let them know where and why you think they are wrong and how they could improve.

You'll have to pardon me, kinman, but whenever someone here writes something like "my estimates", I always shake my head and wonder, "Who are you?".

Why should your estimates be any more reliable than YFull's?

At least some of the guys at YFull are actually geneticists.

rms2
07-30-2017, 08:45 PM
Here is something posted by Larry Walker on the L21 Facebook page that may clear things up a bit.



Iain shows
2934 BC (3638 BC — 2362 BC)

Converting to calendar dates:

At https://www.yfull.com/tree/R-L21/ YFull shows
2400 BC (2800 BC – 1900 BC)
Note that this is not the true “formed” (aka convergence aka coalescence) age. It is the TMRCA of L21’s parent P312 and is used by YFull as the “Formed” age for all first level subclades of P312. For an estimate of the true formed age, you have to go to the parent’s “info”. See next paragraph.

At https://www.yfull.com/branch-info/R-P312/ YFull shows
3260 BC (4815 BC – 1700 BC)
YFull does not give confidence intervals for the branches shown on this report. I have estimated the 95% confidence interval shown above based on the standard deviation of the 100 L21 samples shown. There are many more than 100 L21 samples in YFull, but they show a maximum of 100 on this report for any subclade.

Given the overlaps in confidence intervals shown above, there is no significant difference between the estimate of the age of L21 provided by Iain and those provided by YFull.

MacUalraig
07-30-2017, 09:06 PM
Well, maybe if more people ridiculed their numbers, perhaps they would do what Mitchell suggested: modify their methodology.

For what it's worth, instead of 4400 ybp over and over and over and over again, my estimates for age of formation are as follows (maybe a little too early for some to swallow, but at least they make sense with subclades being younger than their parents):

U152 --- 5100 ybp
L2 --- 4900 ybp
Z49 --- 4700 ybp
Z142 --- 4600 ybp
FGC22963 --- 4400 ybp
FGC22940 --- 4200 ybp

If you look at the SNP counting notes in the info boxes then the levels are all out of step eg the final level actually came out at 4499 which was older than several of the branches above. As their help explains...

' In the situations where the general rule is not followed YFull will add an explanatory note at the bottom of the pop-up table'

eg

'NOTE: Age estimation has been taken from upstream subclade R-Z142, its age estimation is less (4400 < 4145)' to account for this.

Unfortunately if you are going to count SNPs then WYSIWYG. ;-) for example one tester on the bottom level still got 38 SNPs.

What is the methodology behind your figures, so we can compare methods please?

kinman
07-30-2017, 10:12 PM
Just one example. If you look at the information notes for FGC22963 (link below), the ages vary from 3208-4589, except for 6092 for YF07007. And the number of SNPs for YF07007 is likewise way out of line with the others. And yet YF07007 is used to estimate ages of all the clades up through U152 (and P312).

If YF07007 had not been used, the age of FGC22940 would have been about 4000 ybp, not 4400 or 4499 ybp (a difference of 400 or 499 years). I don't know if the number of SNPs for YF07007 is just erroneous, but I certainly would not be using it. It probably doesn't affect the age of P312 very much, but by the time you get down to FGC22940, it looks to be an error of at least 400 years.

https://www.yfull.com/tree/R-FGC22963/

------------------------------------------------------------------------------------------------------------


If you look at the SNP counting notes in the info boxes then the levels are all out of step eg the final level actually came out at 4499 which was older than several of the branches above. As their help explains...

' In the situations where the general rule is not followed YFull will add an explanatory note at the bottom of the pop-up table'

eg

'NOTE: Age estimation has been taken from upstream subclade R-Z142, its age estimation is less (4400 < 4145)' to account for this.

Unfortunately if you are going to count SNPs then WYSIWYG. ;-) for example one tester on the bottom level still got 38 SNPs.

What is the methodology behind your figures, so we can compare methods please?

kinman
08-04-2017, 12:27 PM
I wonder if YFull has considered using a "truncated mean" in such cases.



Just one example. If you look at the information notes for FGC22963 (link below), the ages vary from 3208-4589, except for 6092 for YF07007. And the number of SNPs for YF07007 is likewise way out of line with the others. And yet YF07007 is used to estimate ages of all the clades up through U152 (and P312).

If YF07007 had not been used, the age of FGC22940 would have been about 4000 ybp, not 4400 or 4499 ybp (a difference of 400 or 499 years). I don't know if the number of SNPs for YF07007 is just erroneous, but I certainly would not be using it. It probably doesn't affect the age of P312 very much, but by the time you get down to FGC22940, it looks to be an error of at least 400 years.

https://www.yfull.com/tree/R-FGC22963/

------------------------------------------------------------------------------------------------------------

GailT
08-04-2017, 07:00 PM
With the large number of people who have tested BigY or the FGS full Y, it seems like we should have pretty good estimates of the number of SNPs between major subclades like U152 and L2, and therefore better estimates of their relative ages. For the major subclades, is there a table or chart that shows the complete set of SNPs that separate subclade and its parent clade?

kinman
08-04-2017, 07:45 PM
Yes there is,
And it shows more SNPs than any other tree I've seen for P312. Of course, the drawback is that it only includes the results of those who have sent their results to Alex. But the good news is that it is free and therefore includes a lot more results than YFull. It's called The Big Tree: http://www.ytree.net/


With the large number of people who have tested BigY or the FGS full Y, it seems like we should have pretty good estimates of the number of SNPs between major subclades like U152 and L2, and therefore better estimates of their relative ages. For the major subclades, is there a table or chart that shows the complete set of SNPs that separate subclade and its parent clade?

TigerMW
08-09-2017, 10:01 PM
Yes there is,
And it shows more SNPs than any other tree I've seen for P312. Of course, the drawback is that it only includes the results of those who have sent their results to Alex. But the good news is that it is free and therefore includes a lot more results than YFull. It's called The Big Tree: http://www.ytree.net/
Please do not send Big Y results directly to Alex. That does not get you in the age estimates. If you upload your zipped VCF folders to the appropriate Yahoo group File sections both Alex and Iain will get the results and include them in their analyses. Alex does the Big Tree. Iain does the SNP age estimates.

dkm1987
08-14-2017, 05:53 PM
Mike, what if you had sent directly to Alex a long while back, should we resend to the Yahoo Group now as well so Iain can also utilize them?

MitchellSince1893
08-14-2017, 06:10 PM
Mike, what if you had sent directly to Alex a long while back, should we resend to the Yahoo Group now as well so Iain can also utilize them?

Yes. Alex wasn't given permission to share files sent to him, so you will need to send directly to yahoo group which is a public site.