PDA

View Full Version : Iranian qpAdm modeling



Kurd
07-21-2016, 02:44 AM
For the longest time I had no luck getting decent fits for the Iranian members using qpAdm. However, now that I have the Iranian Neolithic genomes things seem to be improving.

Using Iran Neolithics, project member DMXX is starting to get good fits.



SAMPLE
IRAN NEOL
ANATOLIA NEOL
LEVANT NEOL
ANDRONOVO
HAN
TAIL PROBABILITY
CHISQ


DMXX
53.0%
0.0%
12.5%
27.5%
7.0%
88.4%
1.7





Unfortunately, I still have not been able to obtain a decent fit worth posting for Jesus and NK19191

persian
07-21-2016, 03:33 AM
Hi Kurd, how can I send you my file? I have tested through 23andme, thanks.

Kurd
07-21-2016, 06:42 AM
I tried various combinations using Iran N and Iran Chl with Levant N, to no avail. This is the best I can do so far for Humanist (not a great fit). I was optimistic when I had initially seen such great fits for some of the Kurds using Iran N, but am now starting to wonder whether its genetic contribution was more limited. It is also possible that other members require the use of other base populations, that I have not yet used with Iran N, or perhaps are hitherto not sequenced.



SAMPLE
IRAN RECENT
ANATOLIA NEOL
LEVANT BA
ANDRONOVO
HAN
TAIL PROBABILITY
CHISQ


HUMANIST
57.5%
0.0%
38.9%
0.0%
3.6%
69.0%
3.9

DMXX
07-21-2016, 03:41 PM
Very nice. Broadly in keeping with Davidski's Iranian qpAdm results. Specifically, Iranians being a near two-way mixture of Iran_ChL (with extra Iran_N in some cases) and a steppe-derived component.

Perhaps jesus could be modeled with Iran_N, Iran_ChL, Levant_N, Andronovo and Yoruba, and NK19191 with the same ingroup pops as myself, with Anatolian_N swapped for Iran_N?

Zoran
07-21-2016, 09:24 PM
How do you get this done?

Kurd
07-23-2016, 12:01 AM
Hi Kurd, how can I send you my file? I have tested through 23andme, thanks.

Check your PM

Kurd
07-23-2016, 12:07 AM
Very nice. Broadly in keeping with Davidski's Iranian qpAdm results. Specifically, Iranians being a near two-way mixture of Iran_ChL (with extra Iran_N in some cases) and a steppe-derived component.

Perhaps jesus could be modeled with Iran_N, Iran_ChL, Levant_N, Andronovo and Yoruba, and NK19191 with the same ingroup pops as myself, with Anatolian_N swapped for Iran_N?

Iran N with Iran Chl was not a good fit for NK19191, but Iran N with CHG was better. Therefore, it seems that some individuals can be modeled with good to excellent fits without CHG (yourself and some Kurds), others require CHG in combination with Iran N. I will go back and introduce CHG for some of the individuals that had poor fits with Iran N alone. I will also check to see if your fit improves by trading Anatolia N for CHG.

Kurd
07-23-2016, 11:48 AM
After about a dozen tries, I finally got a decent model for NK19191.



SAMPLE
IRAN CHL
LEVANT NEOL
ONGE
EHG
HAN
TAIL PROBABILITY
CHISQ


NK19191
45.4%
24.6%
11.8%
18.2%
0.00%
90.0%
1.6

DMXX
07-24-2016, 05:21 PM
Iran N with Iran Chl was not a good fit for NK19191, but Iran N with CHG was better. Therefore, it seems that some individuals can be modeled with good to excellent fits without CHG (yourself and some Kurds), others require CHG in combination with Iran N. I will go back and introduce CHG for some of the individuals that had poor fits with Iran N alone. I will also check to see if your fit improves by trading Anatolia N for CHG.

It might be more difficult for NK19191 as he has some recent ancestry from the South Caucasus, which could explain why CHG and Iran_N was better than Iran_N and Iran_ChL. Perhaps CHG and Iran_ChL would work better for him in light of that. Not sure if that combo's been attempted.

Kurd
07-25-2016, 04:53 AM
It might be more difficult for NK19191 as he has some recent ancestry from the South Caucasus, which could explain why CHG and Iran_N was better than Iran_N and Iran_ChL. Perhaps CHG and Iran_ChL would work better for him in light of that. Not sure if that combo's been attempted.

I tried that but it was a failure. The standard errors approached 100%.

Kurd
07-25-2016, 04:56 AM
After about a dozen tries, I finally got a decent model for NK19191.



SAMPLE
IRAN CHL
LEVANT NEOL
ONGE
EHG
HAN
TAIL PROBABILITY
CHISQ


NK19191
45.4%
24.6%
11.8%
18.2%
0.00%
90.0%
1.6





After posting this I noticed that the standard errors were a little unacceptably high on the Iran Chl and Onge sources. So I re-ran removing Han Onge and Levant N, and the standard errors dropped quite a bit.

This 2- way model looks good



SAMPLE
IRAN CHL
EHG
TAIL PROBABILITY
CHISQ


NK19191
83.0%
17.0%
74%
4.3


Standard Errors
6.0%
6.0%







Basically this means that Iran Chl is 83% +/- 6%, and EHG is 17% +/- 6%

Edit: My analysis has shown that Iran Chl is not simply a 2-way mix of CHG and Iran N, as has been proposed elsewhere, but rather is likely admixed with something else in addition to Iran N and CHG.