Page 25 of 29 FirstFirst ... 152324252627 ... LastLast
Results 241 to 250 of 290

Thread: Paleolithic DNA from the Caucasus reveals core of West Eurasian ancestry

  1. #241
    Banned
    Posts
    5,173
    Sex

    With regards to ougroups interacting with your test population outside of your sources (geneflow outgroup-test outside of source-test) and dropping your p-values, Lazaridis and I are on the same page. Here is what they wrote in "The genetic structure of the world’s first farmers" SUPP pg 39

    Several populations cannot be modeled even as WHG+EHG+Mota mixtures (Table S4.8), suggesting that their relationship to the outgroups may be more complex. By dropping some of the outgroups, we may remove populations that violate the topology of Fig. S4.12, that is, they interact with the Test population not via the reference populations
    Three populations (Kostenki14, MA1, and Han), when dropped as outgroups, result in the quadruple (Test, WHG, EHG, Mota) being consistent with 3 streams of ancestry for all (or nearly all in the case of Han) Test populations.......Removal of MA1 and Han suggests interactions between West Eurasia and Upper Paleolithic Siberia and East Asia which we explore in Supplementary Information, section 11
    One of the prime rules in qpAdm is not to use pops in pright that have substantial geneflow with the test pop outside the test pop's source pops.

    I would play with this a little but I don't currently have the IBM genomes.

  2. #242
    Registered Users
    Posts
    2,590
    Sex
    Y-DNA (P)
    R1b-DF27 FGC17112
    mtDNA (M)
    H5a1

    I'll start with this one. The one claiming East Asians are just Onge plus MA1, which I said can't be true because pops that fall on a cline do not have that negative of an effect on the outcome.

    left pops:
    Lahu
    Onge
    MA1

    right pops:
    Mbuti_DG
    Ust_Ishim
    Kostenki14
    Papuan

    numsnps used: 293088

    best coefficients: 0.962 0.038

    std. errors: 0.044 0.044

    fixed pat wt dof chisq tail prob
    00 0 2 1.923 0.382356 0.962 0.038

    And here is with Han and Karitiana, which are on the cline that is supposed to exist, but it breaks down.

    left pops:
    Lahu
    Onge
    MA1

    right pops:
    Mbuti_DG
    Ust_Ishim
    Kostenki14
    Papuan
    Karitiana
    Han

    numsnps used: 270926

    best coefficients: 2.612 -1.612

    std. errors: 0.259 0.259

    fixed pat wt dof chisq tail prob
    00 0 4 402.343 0 2.612 -1.612 infeasible

    I will show the other way tomorrow here. Where the cline is real and is not affected by the pright. So, it tells me another pop in pleft is not really needed or is very close to the ancestral pop. I have to take care of my baby right now.

    I know the geneflow between left and right part. Was told that years ago. There are times it is needed to narrow down deeper ancestry and p lefts that share other deeper ancestry. I also find it useful in seeing if a cline is real or if the p left pop is not a reliable source. F3 is backing the infeasibility of the model if Dzudzuana is like the Anatolians. Age does not matter. Just as f3 EHG Armenian Yamnaya worked great. It is about genetic make-up. I'll discuss it more tomorrow if I have time. Just trust me, it isn't sitting right now. I'll know for sure after I have it, but it seems off.
    Last edited by Chad Rohlfsen; 09-27-2018 at 03:37 AM.

  3. #243
    Banned
    Posts
    5,173
    Sex

    Quote Originally Posted by Chad Rohlfsen View Post
    I'll start with this one. The one claiming East Asians are just Onge plus MA1, which I said can't be true because pops that fall on a cline do not have that negative of an effect on the outcome.

    left pops:
    Lahu
    Onge
    MA1

    right pops:
    Mbuti_DG
    Ust_Ishim
    Kostenki14
    Papuan

    numsnps used: 293088

    best coefficients: 0.962 0.038

    std. errors: 0.044 0.044

    fixed pat wt dof chisq tail prob
    00 0 2 1.923 0.382356 0.962 0.038

    And here is with Han and Karitiana, which are on the cline that is supposed to exist, but it breaks down.

    left pops:
    Lahu
    Onge
    MA1

    right pops:
    Mbuti_DG
    Ust_Ishim
    Kostenki14
    Papuan
    Karitiana
    Han

    numsnps used: 270926

    best coefficients: 2.612 -1.612

    std. errors: 0.259 0.259

    fixed pat wt dof chisq tail prob
    00 0 4 402.343 0 2.612 -1.612 infeasible

    I will show the other way tomorrow here. Where the cline is real and is not affected by the pright. So, it tells me another pop in pleft is not really needed or is very close to the ancestral pop. I have to take care of my baby right now.

    I know the geneflow between left and right part. Was told that years ago. There are times it is needed to narrow down deeper ancestry and p lefts that share other deeper ancestry. I also find it useful in seeing if a cline is real or if the p left pop is not a reliable source. F3 is backing the infeasibility of the model if Dzudzuana is like the Anatolians. Age does not matter. Just as f3 EHG Armenian Yamnaya worked great. It is about genetic make-up. I'll discuss it more tomorrow if I have time. Just trust me, it isn't sitting right now. I'll know for sure after I have it, but it seems off.

    I'm not at all surprised that your model goes to crap when you add Han to your outgroup. This is exactly what I'm referring to in my post on the previous page, and what Lazaridis is referring to in the excerpt from his paper, namely you have interaction between Lahu and Han outside of MA1+Onge --> Lahu. Makes perfect sense.



    EDIT: Just to clarify for those who don't understand what I just wrote. The outgroup, Han is more related to your test population (Lahu) than the sources Onge & MA1 are related to Lahu. This should never be done in qpAdm. Outgroups by definition should be less related to the test pop than the sources are to the test pop.
    Last edited by Kurd; 09-27-2018 at 04:31 AM.

  4. The Following User Says Thank You to Kurd For This Useful Post:

     Pribislav (09-27-2018)

  5. #244
    Registered Users
    Posts
    1,706
    Sex
    Omitted

    Is there any way to tell by a qpadm output which right pop could be interacting with a left pop and thusly trashing a fit?

    Asking because I just reran the Natufian model from last page with a pile more right pops, like I said, to try and break it.
    Couldn't really, but Vestonice16 & Sunghir in right did drop tail prob from 0.85 to 0.29 (not sure which caused the ailment, probably Vestonice16 if there's any correlation with F3 outgroup)

    Left pops: Natufian, Mentese_N, GoyetQ116-1, Taforalt
    Right pops: Barcin_N, Iran_N, Kotias, MA1, Anzick-Kennewick, Kostenki14, Iron_Gates_HG, Ukraine_Mesolithic, Tianyuan, Ust_Ishim, Mota, MbutiDG, South_Africa_2000BP

    Snps used: 185k
    Tail prob: 0.85
    Coefficients: 0.615, 0.108, 0.277
    Std. errors: 0.028, 0.031, 0.028

    All nested models fail.

    Compare to F3 outgroup nmonte...
    Natufian
    "Levant_N" 72.4
    "Taforalt" 15.1
    "GoyetQ116-1" 12.5
    "distance%=0.3779 / distance=0.003779"
    Columns: Barcin_N, Iran_N, Kotias, MA1, ANA, Iron_Gates_HG, EHG, Sunghir, Kostenki14, Ust_Ishim

    Keeping in mind Levant_N basically comes out 78% Anatolian, 7% Iran_N, 15% Taforalt
    Last edited by Kale; 09-27-2018 at 04:51 AM.
    Collection of 14,000 d-stats: Hidden Content Part 2: Hidden Content Part 3: Hidden Content PM me for d-stats, qpadm, qpgraph, or f3-outgroup nmonte models.

  6. #245
    Banned
    Posts
    5,173
    Sex

    Quote Originally Posted by Kale View Post
    Is there any way to tell by a qpadm output which right pop could be interacting with a left pop and thusly trashing a fit?
    Try removing the outgroups one at a time starting with the ones you suspect may be more related to your test pop, than the actual pleft sources are to your test pop. After all an outgroup should be an actual outgroup and not a pop that forms a clade with your test pop.

  7. #246
    Registered Users
    Posts
    1,706
    Sex
    Omitted

    No other way? A single qpadm run on my computersaurus takes 15 minutes.
    Collection of 14,000 d-stats: Hidden Content Part 2: Hidden Content Part 3: Hidden Content PM me for d-stats, qpadm, qpgraph, or f3-outgroup nmonte models.

  8. #247
    Registered Users
    Posts
    546
    Sex
    Ethnicity
    Celto-Germanic, Near East
    Nationality
    Murcan
    Y-DNA (P)
    R1b-U152 (Alsace)
    mtDNA (M)
    T2a1a

    United States of America Palestine Germany Ireland
    Back when the Taforalt genomes were released, David ran some nMontes on present day North Africans. I'd like to call attention back to this nMonte run by rz1706, where he significantly improved the fit for Mozabites by including Pelopponese_N_outlier, which basically belonged to the Anatolia_N clade.

    I can confirm Pelopponese_N improves the fits across all of North Africa-- in some cases greatly so:

    Mozabite (Davids run: 2.4367):
    [1] "distance%=1.6497

    Peloponnese_N_outlier 36.60
    Iberomaurusian 30.60
    Iberia_BA 12.75
    Yoruba 10.95
    Natufian 6.90
    Levant_N 2.05


    Saharawi (David's run = 2.3656)
    [1] "distance%=2.0929"

    Iberomaurusian,36
    Peloponnese_N_o,25.2
    Natufian,10.4
    Iberia_BA,10
    Yoruba,9.4
    Levant_N,8.4
    Ethiopia_4500BP,0.6


    Moroccan (David's run: 1.7158):
    [1] "distance%=1.3121"

    Iberomaurusian,25.8
    Peloponnese_N_o,23.2
    Levant_BA,14.6
    Yoruba,14.4
    Iberia_BA,11.6
    Natufian,7.6
    Seh_Gabi_ChL,1.8
    Levant_N,0.8
    Ethiopia_4500BP,0.2


    Tunisian (David's run: 2.0838):
    [1] "distance%=1.6573"

    Tunisian[/B]
    Peloponnese_N_o,22.8
    Iberomaurusian,20.4
    Levant_BA,15.4
    Iberia_BA,13.4
    Natufian,10.4
    Yoruba,9
    Seh_Gabi_ChL,5
    Ethiopia_4500BP,3
    Levant_N,0.6


    Algerian (David's run: 2.5772):
    [1] "distance%=1.8476"

    Peloponnese_N_o,34.8
    Iberomaurusian,24.6
    Iberia_BA,13.8
    Yoruba,11.2
    Natufian,9.8
    Levant_BA,2.4
    Seh_Gabi_ChL,2
    Ethiopia_4500BP,1.4


    Egyptian (David's run: 2.7927)
    [1] "distance%=1.762"

    Natufian,22.4
    Seh_Gabi_ChL,21.6
    Levant_BA,20
    Peloponnese_N_o,19
    Ethiopia_4500BP,7.6
    Yoruba,4.6
    Iberomaurusian,3.4
    Iberia_BA,1.4

    ----------

    If most of North Africa's Anatolia_N affinity came from migrations from Neolithic/Chl/BA Iberia (as alleged by the original Taforalt paper), nMonte should easily prefer those references over Peloponnese_o. If it was brought by Semitic migrations, they'd choose Levant_N or Levant_BA (in the case of Egyptians, who do prefer Natufian and Levant_BA-- but still score 19% Pelopponese).

    Clearly, Pelopponese_N_o is the best available proxy for the Dzudzuana-like ancestry that was present locally in the Maghreb since the Paleolithic!

    I'd also present the abstract from this hopefully upcoming paper:

    In this study, we present the genomic analysis of two human individuals from a cave site in the area of present-day Morocco which were directly dated to the Medieval period. The samples were processed in a designated ancient DNA lab and the genomic data obtained shows standard patterns of authentic ancient DNA with low levels of contamination. Both individuals – which represent the first ancient genome sequence data from North Africa – do not exhibit particular genetic affinities to modern North Africans or any other present-day population in published genotype data sets despite relatively extensive data has been produced from many areas of Africa. In fact, the most parsimonious way to model them genetically is as two-source admixture between Mediterranean Europeans and Southern Africans.
    Why would this relic population prefer S Euros over the local Maghrebis? Because S Euros (particularly Sardinians) are the closest modern proxy for Dzudzuana-like ancestry, from a principal components standpoint.
    Last edited by K33; 09-27-2018 at 01:13 PM.

  9. The Following 8 Users Say Thank You to K33 For This Useful Post:

     Angoliga (09-27-2018),  beyoku (10-02-2018),  Hando (09-30-2018),  IronHorse (09-27-2018),  Megalophias (09-27-2018),  Michalis Moriopoulos (09-27-2018),  pgbk87 (12-18-2018),  Shamayim (10-20-2018)

  10. #248
    Registered Users
    Posts
    1,332
    Sex
    Omitted

    Quote Originally Posted by Kale View Post
    No other way? A single qpadm run on my computersaurus takes 15 minutes.
    Maybe if you start with the most obvious suspects you're lucky and find it quickly.

  11. The Following User Says Thank You to epoch For This Useful Post:

     Hando (09-30-2018)

  12. #249
    Registered Users
    Posts
    1,706
    Sex
    Omitted

    Quote Originally Posted by K33 View Post
    If most of North Africa's Anatolia_N affinity came from migrations from Neolithic/Chl/BA Iberia (as alleged by the original Taforalt paper), nMonte should easily prefer those references over Peloponnese_o.

    Clearly, Pelopponese_N_o is the best available proxy for the Dzudzuana-like ancestry that was present locally in the Maghreb since the Paleolithic!
    I don't think we need anything that radical to explain it. Could just be that it is actually Anatolian ancestry, that it came from Iberia, but just a region or time without much WHG mixture.
    Collection of 14,000 d-stats: Hidden Content Part 2: Hidden Content Part 3: Hidden Content PM me for d-stats, qpadm, qpgraph, or f3-outgroup nmonte models.

  13. The Following 2 Users Say Thank You to Kale For This Useful Post:

     Hando (09-30-2018),  Michalis Moriopoulos (09-27-2018)

  14. #250
    Registered Users
    Posts
    546
    Sex
    Ethnicity
    Celto-Germanic, Near East
    Nationality
    Murcan
    Y-DNA (P)
    R1b-U152 (Alsace)
    mtDNA (M)
    T2a1a

    United States of America Palestine Germany Ireland
    Quote Originally Posted by Kale View Post
    I don't think we need anything that radical to explain it. Could just be that it is actually Anatolian ancestry, that it came from Iberia, but just a region or time without much WHG mixture.
    I used Iberia_N, Iberia_MN, Iberia_Chl, Iberia_Southwest_CA, and Iberia_BA all as potential source pops.... Maghrebis ALL prefer the Peloponnese outlier, in many cases by a wide margin.

    In my experience with nMonte that's far too consistent and powerful of a preference to simply explain as overfitting...

    Have there been any European Early Farmers modeled as 0% WHG? Maybe some of LBK guys from the very earliest European Neolithic. The date of Iberian gene flow across Gibraltar was inferred as the Copper Age, not the early neolithic. An Anatolia_N-like population with 0% WHG/Villabruna/Magdalenian ancestry in Copper Age Spain seems extremely unlikely IMO. Basques today have higher WHG ancestry than their Spanish and French neighbors.
    Last edited by K33; 09-27-2018 at 05:19 PM.

  15. The Following 2 Users Say Thank You to K33 For This Useful Post:

     Hando (09-30-2018),  Shamayim (10-20-2018)

Page 25 of 29 FirstFirst ... 152324252627 ... LastLast

Similar Threads

  1. Replies: 28
    Last Post: 03-19-2020, 01:07 AM
  2. Replies: 5
    Last Post: 12-11-2019, 09:43 AM
  3. Do Southeast Asians have West Eurasian ancestry?
    By Tsakhur in forum Southeastern
    Replies: 60
    Last Post: 08-20-2019, 06:25 AM
  4. How much ancient West Eurasian ancestry do Amerindians have?
    By Tsakhur in forum Autosomal (auDNA)
    Replies: 4
    Last Post: 09-18-2017, 12:06 PM
  5. Replies: 0
    Last Post: 03-14-2016, 05:53 PM

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •