Global25 automated nMonte for South/Central Asian members

[poi]

Finished the first iteration of the automated nMonte generator. All it needs is any master datasheet like the Global25 datasheet and individual members' scores. It can also be configured to whitelist only certain groups and also ensure certain groups are present in the combination.

For example, for most South/Central Asian members, it is fair to have atleast 1 SouthIndian tribal group, atleast 1 Bronze Age Steppe group, and atleast 1 recently analyzed IVC sample. And it had freedom to pick up to 5 groups like samples from the Turan or Mongolian. Normally, that would take nCr(n:24objects, r:5samples) >50,000 possible combinations per person, so each combination would have to run through nMonte, which would take intense computation. But those whitelisting and restrictions have lowered that to just a few hundred per person.

I have (scaled) coordinates from khana, reza, pnb, kush, jortita, monkeydluffy, and sapporo. Just finished the run for Khana.

Using nMonte3 with default penalty=0.001:

[1] "distance%=1.5556"

Khana

Shahr_I_Sokhta_BA1,38.2
Shahr_I_Sokhta_BA3,36.6
Sintashta_MLBA,19
Mongola,4.2
Kashkarchi_BA,2

---

[1] "distance%=1.5695"

Khana

Shahr_I_Sokhta_BA1,38.8
Shahr_I_Sokhta_BA3,36
Sintashta_MLBA,19.4
Mongola,4.6
Sintashta_MLBA_o2,1.2

---

[1] "distance%=1.6099"

Khana

Shahr_I_Sokhta_BA1,43.8
Sintashta_MLBA,20.6
Shahr_I_Sokhta_BA3,17.2
Paniya,15.4
Mongola,3

---

[1] "distance%=1.609"

Khana

Shahr_I_Sokhta_BA1,39.2
Shahr_I_Sokhta_BA3,36.2
Sintashta_MLBA,19.4
Mongola,3.8
West_Siberia_N,1.4

---

[1] "distance%=1.6376"

Khana

Shahr_I_Sokhta_BA1,40
Shahr_I_Sokhta_BA3,34.6
Sintashta_MLBA,20.4
Mongola,5

---

[1] "distance%=1.7938"

Khana

Shahr_I_Sokhta_BA1,40.8
Shahr_I_Sokhta_BA3,32.6
Zevakinskiy_LBA,14.6
Sintashta_MLBA,9.2
Dai,2.8

---

[1] "distance%=1.7236"

Khana

Gonur2_BA,31.4
Shahr_I_Sokhta_BA1,22.2
Sintashta_MLBA,21
Shahr_I_Sokhta_BA3,16.6
Kusunda,8.8

Results for other members will be available tomorrow, hopefully. My intention is to have realistic groupings for South/Central Asians and also fit(even with the penalty) at most in low 2%. Also, any constructive feedback is appreciated.

[Sapporo]

Looks awesome! Can't wait to see you pump out more results.

[bmoney]

Heres my scaled co-ords Poi

,PC1,PC2,PC3,PC4,PC5,PC6,PC7,PC8,PC9,PC10,PC11,PC1 2,PC13,PC14,PC15,PC16,PC17,PC18,PC19,PC20,PC21,PC2 2,PC23,PC24,PC25
Bmoney_scaled,0.047806,-0.041637,-0.14934,0.097869,-0.072013,0.059404,0.00611,0.017307,0.02986,0.01111 6,-0.007145,-0.001049,-0.005203,-0.00812,0.008007,0.001458,0.001956,-0.001774,-0.002388,0.003377,-0.005241,-0.004699,-0.002711,-0.004579,-0.007784

[poi]

Looks awesome! Can't wait to see you pump out more results.

What are the relevant populations/groups in your mind? Khana's was a very nice fit. Hopefully others have similar without too much reconfiguration of the components. But I don't have to touch anything. The results are prepared by fit ranking. Properly understanding different ancient groups/populations/cultures is where I am most lacking at the moment.

Heres my scaled co-ords Poi

,PC1,PC2,PC3,PC4,PC5,PC6,PC7,PC8,PC9,PC10,PC11,PC1 2,PC13,PC14,PC15,PC16,PC17,PC18,PC19,PC20,PC21,PC2 2,PC23,PC24,PC25
Bmoney_scaled,0.047806,-0.041637,-0.14934,0.097869,-0.072013,0.059404,0.00611,0.017307,0.02986,0.01111 6,-0.007145,-0.001049,-0.005203,-0.00812,0.008007,0.001458,0.001956,-0.001774,-0.002388,0.003377,-0.005241,-0.004699,-0.002711,-0.004579,-0.007784

Cool, started your run. Btw, saw your signature and 2.99% fit, here is something better already after 2 minutes of running your coordinates lol. Note that this is with the default penalty, so the fit will improve without a penalty.

[1] "distance%=2.2424"

bmoney

Shahr_I_Sokhta_BA3,38.4
Shahr_I_Sokhta_BA2,26.2
Paniya,22.6
Sintashta_MLBA,12.2
Kusunda,0.6

---

[1] "distance%=2.3974"

bmoney

Shahr_I_Sokhta_BA3,72
Sintashta_MLBA,12.2
Hajji_Firuz_ChL,8.4
Paniya,7.4

[bmoney]

What are the relevant populations/groups in your mind? Khana's was a very nice fit. Hopefully others have similar without too much reconfiguration of the components. But I don't have to touch anything. The results are prepared by fit ranking. Properly understanding different ancient groups/populations/cultures is where I am most lacking at the moment.



Cool, started your run. Btw, saw your signature and 2.99% fit, here is something better already after 2 minutes of running your coordinates lol. Note that this is with the default penalty, so the fit will improve without a penalty.

Haha awesome, i still think the 2.99 is more realistic though.

SA moderns like Paniya should be removed from the calculator given that they are ~27% Iran N and they received their West Eurasian admixture well after Mehrgarh and as such cannot be a historical population source for admixture (they were receivers rather than donors of admixture)

If we use moderns then it should be non SA pops to capture the overflow, Onge for AASI, Mongola for NE Asian, Dai for SE Asian etc as all SA pops are mixed.

You should also use pops that are closer to SA, Ferghana over Sintashta and pops that are newer in the ancients timeline and therefore more likely to be relevant as a population source.

Mixed pops such as as the Scythians should be removed unless they dramatically improve fit as they soak up E Asian and ANE for example.

We should aim for 2.5 or under with 3 or 4 pops max using ancients from the immediate surroundings of SA to avoid overfitting

[poi]

Haha awesome, i still think the 2.99 is more realistic though.

SA moderns like Paniya should be removed from the calculator given that they are ~27% Iran N and they received their West Eurasian admixture well after Mehrgarh and as such cannot be a historical population source for admixture (they were receivers rather than donors of admixture)

If we use moderns it should be non SA pops to capture the overflow, Onge for AASI, Mongola for NE Asian, Dai for SE Asian etc as all SA pops are mixed.

You should also use pops that are closer to SA, Ferghana over Sintashta and pops that are newer in the ancients timeline and therefore more likely to be relevant as a population source.

Mixed pops such as as the Scythians should be removed unless they dramatically improve fit as they soak up E Asian and ANE for example

I have most those populations white listed, so they should all eventually run overnight. Good point about the modern pops and nearer groups. I will create a filter to remove modern South Asian populations, except for Onge. I even tried adding the Iron Age samples and they ate up both Bronze Age IVC and Steppe.

[jortita]

PC1 PC2 PC3 PC4 PC5 PC6 PC7 PC8 PC9 PC10 PC11 PC12 PC13 PC14 PC15 PC16 PC17 PC18 PC19 PC20 PC21 PC22 PC23 PC24 PC25
Jortita_scaled 0.050082168 -0.124910118 -0.11803877 0.076874368 -0.044315985 0.046853631 0.005640255 0.009922663 0.021679527 0.01640125 -0.009743305 -0.001049066 -0.003865178 -0.003991058 -0.005971693 -0.000530358 0.014863781 -0.005700987 -0.005027922 0.0018759 0.001497358 0.001483833 0.003327688 0.002048482 0.00981947860122929

Thank you

[Reza]

Looking fwd to the results poi. A few more scaled Bengali coordinates in case I've not sent..



,PC1,PC2,PC3,PC4,PC5,PC6,PC7,PC8,PC9,PC10,PC11,PC1 2,PC13,PC14,PC15,PC16,PC17,PC18,PC19,PC20,PC21,PC2 2,PC23,PC24,PC25

Reza_A,0.037562,-0.118817,-0.133124,0.097223,-0.057549,0.063866,0.00141,0.011999,0.038655,0.0180 41,-0.004384,-0.004796,0.003122,0.007156,-0.002307,-0.007425,-0.005085,-0.000633,0.001508,0.002001,-0.001747,0.004575,-0.001109,0.001325,0.00491

Reza_Mother,0.044391,-0.103584,-0.132369,0.093024,-0.050779,0.057451,0,0.009461,0.034769,0.02041,-0.003573,-0.002098,-0.002379,0.00523,-0.000814,-0.016839,-0.020992,0.003167,0.001508,-0.004252,-0.002371,-0.00136,0.00493,0.002048,0.002155

Reza_Wife,0.044391,-0.111708,-0.138026,0.095608,-0.055395,0.056615,-0.00047,0.010615,0.026793,0.020593,0.004547,-0.001349,-0.000446,-0.005092,-0.000814,0.00716,0.011343,-0.002787,-0.002011,0.006003,0.002496,-0.000371,0.006902,0.011447,-0.006227

Tanzil,0.046667,-0.120848,-0.136518,0.093993,-0.052317,0.065539,0.00282,0.015461,0.039882,0.0246 02,-0.003248,0.002698,-0.0055,0.00289,0.001357,0.000928,0.014473,-0.000507,-0.008673,0.004752,0.001872,-0.001484,-0.002465,0.004699,-0.002754

[Sapporo]

What are the relevant populations/groups in your mind? Khana's was a very nice fit. Hopefully others have similar without too much reconfiguration of the components. But I don't have to touch anything. The results are prepared by fit ranking. Properly understanding different ancient groups/populations/cultures is where I am most lacking at the moment.

Most of your populations are fine but these are the top ones (you already have most) imo:

Shahr_I_Sokhta_BA3
Shahr_I_Sokhta_BA2
Parkhai_Eneolithic
Ganj_Dareh_N
Sintashta_MLBA
Srubnaya_MLBA
Hajji_Firuz_ChL
Mongola
Onge


Gonur 2 and Sarazm are other options as well. I might add a SE Asian population too such as Burmese for those around Assam and Bengal. I think they're a little too much South Asian admixed but their geographic proximity makes them a good source. Is it possible to get Tibetans for individuals with Himalayan ancestry?

[bmoney]

I'd say use the Ferghana Kashkarchi_BA R1a z93 sample as its a closer source, instead of Srubnaya and Sintashta MLBA