Global25 automated nMonte for South/Central Asian members

[Sapporo]

@kamil

What are your parameters (I used scaled, no pen, 500 batches/cycles)? When I try to run your neolithic model on Genoplot, no LAO Hoabinhan is being picked up, Karelia_HG is much higher and Tyumen_HG much lower for most.


sample: Gujar Pakistan:Average
distance: 1.7655
Ganj_Dareh_N: 50
Simulated_AASI_NW_by_DMXX: 18.2
Karelia_HG: 16.6
Barcin_N: 11.4
Boshan_N: 2.8
Tyumen_HG: 1
GEO_CHG: 0
Kolyma_Meso: 0

sample: Khatri:Average
distance: 1.735
Ganj_Dareh_N: 46.6
Simulated_AASI_NW_by_DMXX: 17.2
Karelia_HG: 14.6
Barcin_N: 14.2
Boshan_N: 3
Tyumen_HG: 4.4
GEO_CHG: 0
Kolyma_Meso: 0

sample: Kohistani:Average
distance: 1.7459
Ganj_Dareh_N: 46.4
Simulated_AASI_NW_by_DMXX: 16.2
Karelia_HG: 11.8
Barcin_N: 13.6
Boshan_N: 4.8
Tyumen_HG: 7.2
GEO_CHG: 0
Kolyma_Meso: 0

sample: Punjabi JattJ004
distance: 2.3679
Ganj_Dareh_N: 45.6
Simulated_AASI_NW_by_DMXX: 17.2
Karelia_HG: 19.2
Barcin_N: 12.6
Boshan_N: 1.8
Tyumen_HG: 2.6
GEO_CHG: 0
Kolyma_Meso: 1

sample: Kamboj:Average
distance: 1.7192
Ganj_Dareh_N: 45.4
Simulated_AASI_NW_by_DMXX: 15.6
Karelia_HG: 14.4
Barcin_N: 17.4
Boshan_N: 2.6
Tyumen_HG: 4.6
GEO_CHG: 0
Kolyma_Meso: 0

sample: Punjabi Jatt:Average
distance: 2.0449
Ganj_Dareh_N: 45
Simulated_AASI_NW_by_DMXX: 17.2
Karelia_HG: 22
Barcin_N: 13.4
Boshan_N: 2.4
Tyumen_HG: 0
GEO_CHG: 0
Kolyma_Meso: 0

sample: Tarkalani:Average
distance: 2.071
Ganj_Dareh_N: 44.8
Simulated_AASI_NW_by_DMXX: 11.6
Karelia_HG: 14.8
Barcin_N: 18.2
Boshan_N: 3
Tyumen_HG: 5.4
GEO_CHG: 2.2
Kolyma_Meso: 0

sample: Yusufzai:Average
distance: 1.7446
Ganj_Dareh_N: 44.4
Simulated_AASI_NW_by_DMXX: 13
Karelia_HG: 15
Barcin_N: 17.8
Boshan_N: 3.2
Tyumen_HG: 4.8
GEO_CHG: 1.8
Kolyma_Meso: 0

sample: Uthmankhel:Average
distance: 1.9724
Ganj_Dareh_N: 43.4
Simulated_AASI_NW_by_DMXX: 12
Karelia_HG: 14.6
Barcin_N: 18.6
Boshan_N: 3
Tyumen_HG: 6
GEO_CHG: 2.4
Kolyma_Meso: 0

sample: Pashtun:Kandahar2
distance: 2.3199
Ganj_Dareh_N: 40.2
Simulated_AASI_NW_by_DMXX: 8.4
Karelia_HG: 13.6
Barcin_N: 20.6
Boshan_N: 5.6
Tyumen_HG: 7.4
GEO_CHG: 4.2
Kolyma_Meso: 0

sample: Afghan Pashtun:Average
distance: 2.3586
Ganj_Dareh_N: 39.8
Simulated_AASI_NW_by_DMXX: 7.8
Karelia_HG: 17
Barcin_N: 19.8
Boshan_N: 5.8
Tyumen_HG: 2.8
GEO_CHG: 7
Kolyma_Meso: 0

[kamil154]

I used Scaled = On, Penalties = On, Batch/Cycles = 200.

I may or may not use scaled depending on the models, but I always have penalties engaged. Without penalties there is no consistency between consecutive runs even.

[Sapporo]

I used Scaled = On, Penalties = On, Batch/Cycles = 200.

I may or may not use scaled depending on the models, but I always have penalties engaged. Without penalties there is no consistency between consecutive runs even.

I tried using those parameters but the fits were poor and models looked off:


sample: Gujar Pakistan:Average
distance: 3.6043
Ganj_Dareh_N: 53
Simulated_AASI_NW_by_DMXX: 12
Barcin_N: 9.5
LAO_Hoabinhian: 9
Tyumen_HG: 8
Karelia_HG: 6
GEO_CHG: 1.5
Boshan_N: 0.5
Kolyma_Meso: 0.5

sample: Kohistani:Average
distance: 3.1175
Ganj_Dareh_N: 44.5
Simulated_AASI_NW_by_DMXX: 11
Barcin_N: 9
LAO_Hoabinhian: 11
Tyumen_HG: 9
Karelia_HG: 6.5
GEO_CHG: 7
Boshan_N: 1.5
Kolyma_Meso: 0.5

sample: Kamboj:Average
distance: 3.1304
Ganj_Dareh_N: 44.5
Simulated_AASI_NW_by_DMXX: 11
Barcin_N: 14
LAO_Hoabinhian: 8.5
Tyumen_HG: 6
Karelia_HG: 8
GEO_CHG: 6.5
Boshan_N: 0.5
Kolyma_Meso: 1

sample: Yusufzai:Average
distance: 3.2965
Ganj_Dareh_N: 44
Simulated_AASI_NW_by_DMXX: 8
Barcin_N: 15
LAO_Hoabinhian: 7.5
Tyumen_HG: 6.5
Karelia_HG: 8.5
GEO_CHG: 8
Boshan_N: 2
Kolyma_Meso: 0.5

Did you use population averages or individual samples for the Neolithic reference pops?

[kamil154]

Population averages. Your results seem off from mine. NW South Asians don't show up with that much Hoabinhian even when using AASI NW in my runs.

[kamil154]

Attachment 42751

Attachment 42752

[goingmerry]

There clearly has been much gene flow among Austroasiatic-speaking tribes and Dravidian-speaking tribes in Chhattisgarh, Odisha, and coastal Andhra, but the Konda Dora do not speak an Austroasiatic language. Their tribal language is related to the languages of the Khonds and Gonds, so they are Dravidian from a linguistic perspective.

Makes sense, I'm a Telugu with roots Coastal Andhra. Doing some research on my mtDNA, M40a, I found that it originates in Eastern India and is one of the Munda specific mtDNA haplogroups. Another study I found that it also appears amongst Musahars, a tribal turned Dalit community in Bihar.

Link to the Munda study: https://www.researchgate.net/publica...ation_in_India

[kamil154]

What type of model do you reckon is more plausible for most NW South Asians and why?

Using a lower Iran N, CHG+ANF rich BMAC proxy such as Gonur3 or Sappali Tepe2, frees up space for a more Iran N shifted IVC base. At the same time, using a more ASI shifted IVC proxy than necessary (BA3 in my case), tends to overinflate BMAC to compensate.

sample: kamil154
distance: 2.3473
Shahr_I_Sokhta_BA2: 60.5
Srubnaya_Alakul_MLBA: 22
Gonur3_BA: 13.5
Chokhopani_2700BP: 4

VS

sample: kamil154
distance: 2.353
Shahr_I_Sokhta_BA2: 49.5
Gonur1_BA: 28.5
CG_CentralSteppeMLBA: 17.5
Chokhopani_2700BP: 4.5

VS

sample: kamil154
distance: 2.3632
Gonur1_BA: 47
Shahr_I_Sokhta_BA3: 35
Krasnoyarsk_MLBA: 14
Chokhopani_2700BP: 4

Not sure what to believe in at this point. But my sense is to go with the model with which a greater percentage of the ancestry can be represented using a more indigenous IVCp input.

[pegasus]

What type of model do you reckon is more plausible for most NW South Asians and why?

Using a lower Iran N, CHG+ANF rich BMAC proxy such as Gonur3 or Sappali Tepe2, frees up more space for a more Iran N shifted IVC base. At the same time using a more ASI shifted IVC proxy than necessary (BA3 in my case), tends to overinflate BMAC to compensate.

sample: kamil154
distance: 2.3473
Shahr_I_Sokhta_BA2: 60.5
Srubnaya_Alakul_MLBA: 22
Gonur3_BA: 13.5
Chokhopani_2700BP: 4

VS

sample: kamil154
distance: 2.353
Shahr_I_Sokhta_BA2: 49.5
Gonur1_BA: 28.5
CG_CentralSteppeMLBA: 17.5
Chokhopani_2700BP: 4.5

VS

sample: kamil154
distance: 2.3632
Gonur1_BA: 47
Shahr_I_Sokhta_BA3: 35
Krasnoyarsk_MLBA: 14
Chokhopani_2700BP: 4

Not sure what to believe in at this point. But my sense is to go with the model with which a greater percentage of the ancestry can be represented using a more indigenous IVCp input.

That one is very easy to figure out. If you look at Z scores and gendstats there is a relative affinity for CHG not PPNB among most NW South Asians , and utilizing very Iran Chl shifted sources like Seh Gabi or the samples you utilized like Gonur3 will drive models to failure, you even see that with Vagheesh's models for modern populations as well. What you are implying is heavily Iran Chl types were the main population in Central Asia in the Bronze Age, thats not the case and especially in rural settings where I imagine people would generally have less Iran Chl type ancestry compared with those in cities. The scenario your suggesting would be true though during the Achaemenid expansion and you blatantly see it with those Khwarzemian outliers from Mongolia. Though by this juncture, Indo Aryans are long gone. Come to think of it explains why some modern samples prefer Parkhai MBA, I remember Yamamoto even mentioning this. Obviously some modern populations will have accrued other ancestries but if your talking about NW South Asian populations as whole like that Punjabi Sikh cluster, definitely not.


Wrt , IVC samples , the paradox is they have AASI levels like modern day populations but everything else is quite different. Even with the low Steppe SGPT outliers , when you add 1456,1466,8726, the p value falls like a rock because they have too much Iran_N and not the other varied components Indo Aryans or closely related populations are bringing in.

[vishankar]

Makes sense, I'm a Telugu with roots Coastal Andhra. Doing some research on my mtDNA, M40a, I found that it originates in Eastern India and is one of the Munda specific mtDNA haplogroups. Another study I found that it also appears amongst Musahars, a tribal turned Dalit community in Bihar.

Link to the Munda study: https://www.researchgate.net/publica...ation_in_India

can you post your nmonte runs?....or harappa scores??

[goingmerry]

can you post your nmonte runs?....or harappa scores??

Here's my Harappa score. I'm interested in getting a nMonte but I don't know where to start, still new to understanding South Asian genetics. If somebody could link to a guide, that would be much appreciated.

Admix Results (sorted):

# Population Percent
1 S-Indian 54.97
2 Baloch 33.09
3 Caucasian 6.96
4 Siberian 1.88
5 Papuan 1.61
6 NE-Asian 1.04
7 SE-Asian 0.23
8 San 0.21

Single Population Sharing:

# Population (source) Distance
1 karnataka (harappa) 4.02
2 tamil (harappa) 4.09
3 andhra-pradesh (harappa) 4.29
4 velama (reich) 4.31
5 ap-hyderabad (harappa) 4.32
6 sinhalese (harappa) 5.25
7 velama (metspalu) 5.25
8 ap-reddy (harappa) 5.36
9 tamil-vellalar (harappa) 5.75
10 tharu (metspalu) 5.83
11 tamil-nadu-scheduled-caste (metspalu) 6.01
12 kerala-muslim (harappa) 6.44
13 tamil-nadar (harappa) 6.67
14 up-muslim (metspalu) 6.77
15 kerala-christian (harappa) 6.79
16 kurumba (metspalu) 6.85
17 singapore-indian-b (sgvp) 7.05
18 lodi (reich) 7.48
19 caribbean-indian (harappa) 7.84
20 dharkar (metspalu) 7.85

Mixed Mode Population Sharing:

# Primary Population (source) Secondary Population (source) Distance
1 93.4% tamil-vellalar (harappa) + 6.6% abhkasian (yunusbayev) @ 2.03
2 93.5% tamil-vellalar (harappa) + 6.5% georgian (behar) @ 2.05
3 92.9% tamil-vellalar (harappa) + 7.1% north-ossetian (yunusbayev) @ 2.27
4 92.9% tamil-vellalar (harappa) + 7.1% adygei (hgdp) @ 2.34
5 92.9% tamil-vellalar (harappa) + 7.1% balkar (yunusbayev) @ 2.39
6 92.8% tamil-vellalar (harappa) + 7.2% chechen (yunusbayev) @ 2.56
7 92.6% tamil-vellalar (harappa) + 7.4% kumyk (yunusbayev) @ 2.56
8 93.1% tamil-vellalar (harappa) + 6.9% armenian (behar) @ 2.59
9 92.3% tamil-vellalar (harappa) + 7.7% azeri (harappa) @ 2.63
10 92.6% tamil-vellalar (harappa) + 7.4% lezgin (behar) @ 2.66
11 95.9% velama (reich) + 4.1% georgian (behar) @ 2.73
12 92.3% tamil-vellalar (harappa) + 7.7% kurd (yunusbayev) @ 2.73
13 95.8% velama (reich) + 4.2% abhkasian (yunusbayev) @ 2.75
14 91.8% tamil-vellalar (harappa) + 8.2% iranian (harappa) @ 2.75
15 95.4% velama (reich) + 4.6% balkar (yunusbayev) @ 2.75
16 95.4% velama (reich) + 4.6% north-ossetian (yunusbayev) @ 2.75
17 94.9% velama (reich) + 5.1% nogai (yunusbayev) @ 2.77
18 92.4% tamil-vellalar (harappa) + 7.6% kurd (xing) @ 2.78
19 50% velama (reich) + 50% ap-hyderabad (harappa) @ 2.79
20 95.5% velama (reich) + 4.5% adygei (hgdp) @ 2.79