Page 1 of 4 123 ... LastLast
Results 1 to 10 of 40

Thread: Continuum Population Modeling using Vahaduo - My first attempt :)

  1. #1
    Registered Users
    Posts
    866
    Ethnicity
    Greek
    Nationality
    American

    United States of America Greece

    Continuum Population Modeling using Vahaduo - My first attempt :)

    So, after a couple of months of posting a little, and reading a lot, I want to try my hand at some modeling. I want to thank everyone who participates in modeling and discussions of models, and a special thanks to Agamemnon and Michalis, as this model is based mostly on their previous examples.

    Tools used: VahaduoJS 19.08, Global_25_PCA_pop_averages_scaled, my own scaled G25 coordinates

    Purpose: Continuum populations look very similar on many PCAs, including early Eurogene calculators like K15. The purpose of this model is to give the software specific choices of possible ancestry (like Slav vs. Celt, or Levant vs. North Africa) to find the differences between these populations. Fit does matter, but since there are many target populations, the "average fit" for the matrix was what I am looking at during the first run. This model is not specific to any one of these populations. Again, I am trying to look at the continuum populations as a whole.

    Source Samples - I tried to use Metal Age samples, Bronze Age if possible. As you will see I didn't entirely succeed at this, I will try to explain why.

    GRC_Mycenaean and Iberia_Northeast_Empuries2 - Following the examples of Aga and Michalis, I used two ancient Greek samples to try and capture the early Greek diversity on both sides of the Aegean Sea. Trying to use only one of the two drastically decreases fit, so both are being use. Also considered: Anatolia_Isparta_EBA. This samples worked well too, but fit was slightly better for Empuries_2

    CZE_Early_Slav - Seemingly the standard model choice for the Slavic Migrations.

    ARM_LBA - Following Aga again on this one. This sample only appears in small percents, but greatly helps fit. I assume it accounts for any extra Caucasus like ancestry.

    ITA_Collegno_MA - I wanted something metal age out of Italy that was unlikely to have too much Continuum-like ancestry. In this model, it is offered as a "choice" to CZE_Early_Slav to account for Steepe like ancestry. Also considered: Any of the Bell Beaker populations. In the end, I went with this one because it is specific to Italy.

    Levant_Canaanite_MBA - I found picking the right Levant population difficult. In the end, this one seems to overlap the Greek samples less than Levant_LBN_Roman in this model, so it was chosen. Also considered: Levant_LBN_Roman - this produced nearly identical fits. I thought of adding both, but I feared the "overfit" if this sample eats into to the Greek and other Levant samples.

    Canary_Islands_Guanche - After reading other conversations, I thought a North African population was needed for both fit and historical purposes. I hope this is an ancient sample.....pretty sure it is, but I could be wrong. Also considered: EGY_Late_Period and EGY_Hellenistic - These were really fun to use, and created good fits, but in the end didn't seem like realistic options based on history. The Egyptian samples cut into the Levant amounts by alot and the Sicilian and Jewish target populations all ended up with 10-21%, which I didn't think was as accurate as the 3-8% range the Canary Island sample produced.

    Target Samples -

    Greeks - Cypriot, Greek_Crete, GreekScholar_scaled, - Not much choice here. I picked the two Greek populations from the islands, and included myself as a third sample. My ancestry is from the NE_Aegean islands, mostly Fourni.
    Italians - Italian_Abruzzo, East_Sicily, West_Sicily - I picked Abruzzo as the counter to Cypriot, more or less the 'western' and 'eastern' ends of the Continuum populations. I went with both Sicily samples because they are frequent conversation topics here, and I have experience charting them on the K15 PCA.
    Jewish- Ashkenazi_Jew, Italian_Jew, Romaniote_Jew - Ashkenazi is another much talked about population, lots of models produced and discussed. I thought picking Italian and Greek Jews would be interesting as well since you can compare them both to the Ashkenazi sample, and their gentile neighbors in Italy and Greece.

    So, this is my first model, and I am sure I messed lots of stuff up. Please give me any feedback you have about any of it, the sources, the purpose, the targets, etc. I can also use these same source populations for other target populations as well. Just let me know what you want to see ��

    Conclusions: Well, I don't want to form any until I get some feedback on the model. I would say the ITA_Collengo_MA vs. CZE_Early_Slave results are interesting.

    Last edited by Greekscholar; 09-23-2019 at 02:47 AM.

  2. The Following 4 Users Say Thank You to Greekscholar For This Useful Post:

     Agamemnon (09-23-2019),  Cascio (10-30-2019),  jonahst (09-23-2019),  Targum (09-23-2019)

  3. #2
    Registered Users
    Posts
    866
    Ethnicity
    Greek
    Nationality
    American

    United States of America Greece
    Whoops. I think ITA_Collegno_MA is "Middle Ages" not "Metal Age." I swapped it out for Bell_Beaker_ITA. Fit improves, but Greek sources crash hard for non-Greek target populations, seemingly replaced by ARM_LBA. This is probably wrong, isn't it?


    Last edited by Greekscholar; 09-23-2019 at 02:51 AM.

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

     Agamemnon (09-23-2019)

  5. #3
    Banned
    Posts
    4,342
    Sex
    Location
    USA
    Nationality
    American
    Y-DNA (P)
    R1a1a
    mtDNA (M)
    H

    United States of America
    It is interesting how even though all of these populations roughly plot together, when you break down their components they are really quite different.

    You are significantly more Slavic than the Cretan sample. I think this is because North Aegean islanders may have received more migration from the mainland than what you see in Crete. Therefore I would imagine a northern mainland Greek sample would have Slavic approaching 40%, because they should be more Slavic than the islands. But even still, the Slavic input is one of the factors differentiating even Crete from the South Italians. The other being an Italic-like component captured by the Collegno sample that is not present in Crete.

    For clarity to those who haven't read my previous posts... West Sicily is Trapani, East Sicily is Ragusa, both being academic samples. Neither sample, nor the Abruzzese, come up with any significant Slavic signal.

    Trapani is probably the least Hellenized part of South Italy, and second to Malta is the area with the next highest North African. It's interesting that both Sicilian samples have roughly equal Levantine, but the West sample has higher North African, which suggests to me that these migrations likely did not happen at the same time.. and if you add up Levant + North African, it'll almost always come out highest in western Sicily.

    Cyprus is, far and away, more Levantine than any of the other Continuum samples which is why they are transitional. You can say that Sicilians, Cretans, Western Jews still plot as "European" even if on the edge of that cluster but Cypriots are transition.

    On the note of why the West Sicily sample is so different, see this from the Crete study. All of the Sicilian samples -- Palermo, Ragusa, Catania, Agrigento -- plot right with Crete while the Trapanese deviate, and I believe this is because they are significantly less ancient Aegean.

    This is a direct quote from the study's Supplementary Information that I think is consistent with what you found.

    PCA plot for Cretans and Sicily. There is an overlap between most of the Sicilians with the Cretan samples. With the exception of two samples from Trapani, the Sicilians seem to be in the same cluster as the Cretans.


    Quote Originally Posted by Greekscholar View Post
    Whoops. I think ITA_Collegno_MA is "Middle Ages" not "Metal Age." I swapped it out for Bell_Beaker_ITA. Fit improves, but Greek sources crash hard for non-Greek target populations, seemingly replaced by ARM_LBA. This is probably wrong, isn't it?



    Ok now you are getting what I did -- Trapani as a mixture of Levantine, Berber, Italic, and North European with almost no Greek. But you still need Greek for the other two south Italian samples.

    And you also found Ashkenazim not to have much Greek-like admixture, and to be more Italic. This is what I found too!
    Last edited by Sikeliot; 09-23-2019 at 03:11 AM.

  6. The Following 5 Users Say Thank You to Sikeliot For This Useful Post:

     Agamemnon (09-23-2019),  Cascio (10-30-2019),  Greekscholar (09-23-2019),  jonahst (09-23-2019),  Targum (09-23-2019)

  7. #4
    Banned
    Posts
    2,951
    Sex
    Location
    America
    Ethnicity
    Telgu/marathi

    Siberian Tatars Tajikistan
    Some of the components are already quite similar to one another, that allows some populations to cluster next to one another despite getting drastically different amounts of one component or another.

  8. #5
    Registered Users
    Posts
    866
    Ethnicity
    Greek
    Nationality
    American

    United States of America Greece
    Thanks for the analysis. I think you hit on the big point this model shows, Jewish people and Italians both show at least some North African and some version of Bell Beaker-like Steppe ancestry. The Greek samples do not, they have Slavic and Levantine instead. I will duplicate the source populations and some some new target populations, including Mainland Greece tomorrow. I had Mainland Greece in my pilot runs, but I tried to keep this first role out to 9 targets.....which is already a lot to digest.

    Anybody have any guess on why the Bell Beaker inclusion pushed some of much of the Mycenaean amount towards Armenia? The fit may be closer, but I am not sure that model is a reflection of history.

  9. The Following 2 Users Say Thank You to Greekscholar For This Useful Post:

     Agamemnon (09-23-2019),  Targum (09-23-2019)

  10. #6
    Banned
    Posts
    4,342
    Sex
    Location
    USA
    Nationality
    American
    Y-DNA (P)
    R1a1a
    mtDNA (M)
    H

    United States of America
    Out of curiosity can you show me what it looks like when you include both Egypt and Levant but exclude Guanche?

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

     Greekscholar (09-23-2019)

  12. #7
    Banned
    Posts
    4,342
    Sex
    Location
    USA
    Nationality
    American
    Y-DNA (P)
    R1a1a
    mtDNA (M)
    H

    United States of America
    Quote Originally Posted by Greekscholar View Post
    Thanks for the analysis. I think you hit on the big point this model shows, Jewish people and Italians both show at least some North African and some version of Bell Beaker-like Steppe ancestry. The Greek samples do not, they have Slavic and Levantine instead. I will duplicate the source populations and some some new target populations, including Mainland Greece tomorrow. I had Mainland Greece in my pilot runs, but I tried to keep this first role out to 9 targets.....which is already a lot to digest.

    Anybody have any guess on why the Bell Beaker inclusion pushed some of much of the Mycenaean amount towards Armenia? The fit may be closer, but I am not sure that model is a reflection of history.
    All of the Continuum populations show Levantine admixture (I think this will fall sharply for mainland Greeks), the difference seems to be Slavic versus Italic, and the lack of North African in the Greeks.

    The inclusion of Bell Beaker has the sharpest impact on the Trapanese and Ashkenazim, who end up losing their Mycenaean in favor of Armenian/Caucasian, which I noticed the exact same thing when I tried modeling them. This suggests to me that both groups have very little Greek ancestry, but did acquire Anatolian or Caucasian admixture from a different migration -- maybe Elymians for Trapani, and the migration of Jews into Anatolia for Ashkenazim.

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

     Greekscholar (09-23-2019),  Targum (09-23-2019)

  14. #8
    Registered Users
    Posts
    866
    Ethnicity
    Greek
    Nationality
    American

    United States of America Greece
    Quote Originally Posted by Censored View Post
    Some of the components are already quite similar to one another, that allows some populations to cluster next to one another despite getting drastically different amounts of one component or another.
    Yes, some of that is on purpose to try and force the software to make a choice (Slav vs. Bell Beaker?) (Levant vs. North Africa?) The two ancient Greek samples are an attempt to "capture" as much ancient Greek ancestry as possible, so I am adding both together rather than viewing the results as one or the other. Including both really reduces distance though.......probably the dreaded "overfit" I read so much about.

    Any suggestions on how to improve the model? This is my first one.

  15. #9
    Registered Users
    Posts
    866
    Ethnicity
    Greek
    Nationality
    American

    United States of America Greece
    Quote Originally Posted by Sikeliot View Post
    Out of curiosity can you show me what it looks like when you include both Egypt and Levant but exclude Guanche?
    You bet, both Egyptian samples (Late and Hellenistic) take a big chunk out of the Levant totals, which is ultimately why I went with Guanche instead.

  16. #10
    Banned
    Posts
    4,342
    Sex
    Location
    USA
    Nationality
    American
    Y-DNA (P)
    R1a1a
    mtDNA (M)
    H

    United States of America
    Quote Originally Posted by Greekscholar View Post
    You bet, both Egyptian samples (Late and Hellenistic) take a big chunk out of the Levant totals, which is ultimately why I went with Guanche instead.
    I actually believe the primary source of North African admixture into Sicily is from Egypt and Tunisia, and therefore using Egypt is probably more historically accurate for them than using Guanches. Most far NW African migration into Europe went into Iberia instead. While Sicilian Arabic/Maltese is a Maghrebi dialect, it is possible more actual migration occurred from Egypt.

    I wonder if forced to choose between Egypt and Levant, if Sicilians and Jews will lose more of their Levantine to Egyptian than Greek islanders would. I genuinely believe there was very, very little North African input into Greece.

    What did you think of the PCA plot from the Crete study? Do you think it explains what we see in your models at least partially?

  17. The Following User Says Thank You to Sikeliot For This Useful Post:

     eolien (09-23-2019)

Page 1 of 4 123 ... LastLast

Similar Threads

  1. Replies: 1
    Last Post: 10-02-2017, 12:54 AM
  2. Replies: 2
    Last Post: 01-25-2017, 11:27 AM
  3. Replies: 3
    Last Post: 03-04-2016, 06:07 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
  •