Page 3 of 14 FirstFirst 1234513 ... LastLast
Results 21 to 30 of 132

Thread: Mephisto's Grand K12 Modern Calculator

  1. #21
    Gold Member Class
    Posts
    1,898
    Sex
    Location
    Pennsylvania
    Ethnicity
    West European
    Nationality
    USA
    Y-DNA
    R1b-U152-Z36+FGC6511
    mtDNA
    H11a2a

    United States of America Germany England France Scotland Ireland
    Quote Originally Posted by mephisto View Post
    Yes, michal, however, he even knows that he has SE Asian ancestry at around 5,25% (as his signature states), so this comes clearly from that (this is not even specualtion in this case, count E_Asian, Yakut and Papuan together and you get 4,41%, that is as close as a calculator can get ).
    I think he was talking about my father's 2.5+ % score. This is usual for him along with other Asian and some African, but I haven't yet found the ancestral source. I am recently getting some good hints.

  2. The Following 3 Users Say Thank You to randwulf For This Useful Post:

     mephisto (01-26-2018),  michal3141 (01-26-2018),  Pylsteen (01-26-2018)

  3. #22
    Banned
    Posts
    343
    Sex
    Y-DNA
    R1b

    Bad news, ADMIXTURE lumped East and NW Europe together in one big component, Mordovians score 90% of it, Orcadians and French score around 70% of it while the rest is Mediterranean (Sardinian). Does it make sense to release this "updated" version or should I try to remove some south Asian populations (like Kalash) and re-run it? I ran it as K13.
    Last edited by mephisto; 01-26-2018 at 11:28 PM.

  4. #23
    Registered Users
    Posts
    710
    Sex
    Location
    Krakow
    Ethnicity
    Polish
    Nationality
    Pole
    Y-DNA
    R1a > M458 > Y23110
    mtDNA
    J1c1

    Poland
    Quote Originally Posted by mephisto View Post
    Bad news, ADMIXTURE lumped East and NW Europe together in one big component, Mordovians score 90% of it, Orcadians and French score around 70% of it while the rest is Mediterranean (Sardinian). Does it make sense to release this "updated" version or should I try to remove some south Asian populations (like Kalash) and re-run it?
    You are running unsupervised admixture? Maybe try with supervised as we (me and LukaszM) are doing.

  5. The Following User Says Thank You to michal3141 For This Useful Post:

     mephisto (01-26-2018)

  6. #24
    Banned
    Posts
    6,337
    Sex
    Location
    Torun
    Ethnicity
    Central 75% + 25% Mazovia
    Nationality
    Pole
    Y-DNA
    R1a > M198 > YP 1337

    Poland
    Try also with Balts instead of Mordovian. When I run low K unspervised calcs, Balts took over NE-Euro component nearly all the time (but I have few thousands samples in dataset...).

  7. The Following User Says Thank You to lukaszM For This Useful Post:

     mephisto (01-26-2018)

  8. #25
    Banned
    Posts
    343
    Sex
    Y-DNA
    R1b

    Quote Originally Posted by michal3141 View Post
    You are running unsupervised admixture? Maybe try with supervised as we (me and LukaszM) are doing.
    Yes, it is unsupervised. I made bad experiences in the last days with supervised runs (if it is that what I assume) of my own. What do you exactly mean with supervised? Do you mean that you merge all references for one ancestral component together to one big .bed file and then edit the .fam file (for example: 1 E_Europe [this .bed file would include all my Mordovian samples for example] 0 0 0 0) so that the file is treated as one big reference (I hope you understand what I mean), or how do you do a supervised run? With my own experiment my example candidate ended up scoring 7% West Africa which was wrong.
    Last edited by mephisto; 01-26-2018 at 11:40 PM.

  9. #26
    Registered Users
    Posts
    710
    Sex
    Location
    Krakow
    Ethnicity
    Polish
    Nationality
    Pole
    Y-DNA
    R1a > M458 > Y23110
    mtDNA
    J1c1

    Poland
    Quote Originally Posted by mephisto View Post
    Yes, it is unsupervised. I made bad experiences in the last days with supervised runs of my own. What do you exactly mean with supervised? Do you mean that you merge all references for one ancestral component together to one big .bed file and then edit the .fam file (for example: 1 E_Europe 0 0 0 0) so that the file is treated as one big reference (I hope you understand what I mean), or how do you do a supervised run?
    I am doing this differently. I create a .pop file. Each line in this file is a component name. Let me give you an example. In your .fam file you have the following individuals:
    AFSD (Orcadian)
    31242 (Mordovian)
    33234r3 (Sardinian)
    3224 (Orcadian)
    .....

    Of course .fam file does not contain ethnicity info so you need to record the info about the ethnicity somehow.
    In the .pop file you basically have:
    Orcadian
    Mordovian
    Sardinian
    Orcadian
    ....

    You see that the components names in the .pop file correspond with the ethnicity of individuals in your .fam file?

    This is basicall what you need to do and the only add --supervised when running ADMIXTURE e.g.

    admixture --supervised Mephisto.bed 13

    EDIT: I see you may have many ethnicities for one component. In this case you need to have the names of your components in a .pop file instead of ethnicities of your individuals.

    The bad thing about the supervised run however might be that your reference individuals (samples) will score almost 100% of their respective component. This I personally don't like.
    Last edited by michal3141; 01-26-2018 at 11:45 PM.

  10. The Following User Says Thank You to michal3141 For This Useful Post:

     mephisto (01-26-2018)

  11. #27
    Banned
    Posts
    6,337
    Sex
    Location
    Torun
    Ethnicity
    Central 75% + 25% Mazovia
    Nationality
    Pole
    Y-DNA
    R1a > M198 > YP 1337

    Poland
    Too many new calcs these days

  12. #28
    Banned
    Posts
    343
    Sex
    Y-DNA
    R1b

    Quote Originally Posted by michal3141 View Post
    I am doing this differently. I create a .pop file. Each line in this file is a component name. Let me give you an example. In your .fam file you have the following individuals:
    AFSD (Orcadian)
    31242 (Mordovian)
    33234r3 (Sardinian)
    3224 (Orcadian)
    .....

    Of course .fam file does not contain ethnicity info so you need to record the info about the ethnicity somehow.
    In the .pop file you basically have:
    Orcadian
    Mordovian
    Sardinian
    Orcadian
    ....

    You see that the components names in the .pop file correspond with the ethnicity of individuals in your .fam file?

    This is basicall what you need to do and the only add --supervised when running ADMIXTURE e.g.

    admixture --supervised Mephisto.bed 13
    Thank you. I did not even know that it was possible But that is a lot of work with 649 samples (actually more since I added the Mordovians).. Is there a "tool" with which you can extract the information maybe out of the .ind file and make it automatically?
    Last edited by mephisto; 01-26-2018 at 11:52 PM.

  13. #29
    Registered Users
    Posts
    710
    Sex
    Location
    Krakow
    Ethnicity
    Polish
    Nationality
    Pole
    Y-DNA
    R1a > M458 > Y23110
    mtDNA
    J1c1

    Poland
    Quote Originally Posted by mephisto View Post
    Thank you. I did not even know that it was possible I have to read myself into it and try.. But that is a lot of work with 649 samples.. Is there a "tool" with which you can extract the information maybe out of the .ind file and make it automatically?
    I don't know. Maybe plink can do this. Personally I do the following process:
    1. Convert .ind (and .geno) to .fam, .bim and .bed.
    2. I do the filtering of my individuals in plink on .bed, .fam and .bim
    3. I keep the list of my selected individuals and based on their IDs I filter the .ind file to get the list of sorted ethnicities!
    4. I convert this list of sorted ethnicities into component names. You can write a simple script in Python or other programming language to map your ethnicities into component names. This is how I get .pop file

    I admit it requires some work. Maybe there is an easier way by using magic of plink

    The thing about NE-Euro and NW-Euro basically collapsing into one component is also what I saw in my first calc. Interestingly Mordovians where getting like 95% of this one common component and English people were getting about 60-70% and the rest was Mediterranean for them. Interesting you saw the same behavior.
    Last edited by michal3141; 01-27-2018 at 12:03 AM.

  14. The Following User Says Thank You to michal3141 For This Useful Post:

     mephisto (01-27-2018)

  15. #30
    Banned
    Posts
    6,337
    Sex
    Location
    Torun
    Ethnicity
    Central 75% + 25% Mazovia
    Nationality
    Pole
    Y-DNA
    R1a > M198 > YP 1337

    Poland
    Quote Originally Posted by michal3141 View Post
    I don't know. Maybe plink can do this. Personally I do the following process:
    1. Convert .ind (and .geno) to .fam, .bim and .bed.
    2. I do the filtering of my individuals in plink on .bed, .fam and .bim
    3. I keep the list of my selected individuals and based on their IDs I filter the .ind file to get the list of sorted ethnicities!
    4. I convert this list of sorted ethnicities into component names. You can write a simple script in Python or other programming language to map your ethnicities into component names. This is how I get .pop file

    I admit it requires some work. Maybe there is an easier way by using magic of plink
    Yes, two columns in excel And magic of mouse copying...
    Last edited by lukaszM; 01-27-2018 at 12:04 AM.

  16. The Following 2 Users Say Thank You to lukaszM For This Useful Post:

     mephisto (01-27-2018),  michal3141 (01-27-2018)

Page 3 of 14 FirstFirst 1234513 ... LastLast

Similar Threads

  1. MDLP K16 Modern calculator
    By DMXX in forum Autosomal (auDNA)
    Replies: 612
    Last Post: Today, 04:22 AM
  2. Mephisto's K10 Ancient Calculator
    By mephisto in forum Autosomal (auDNA)
    Replies: 21
    Last Post: 02-10-2018, 11:15 PM
  3. Mephisto's K15 Ancient Calculator
    By mephisto in forum Autosomal (auDNA)
    Replies: 88
    Last Post: 01-17-2018, 09:05 PM
  4. Mephisto´s [First] K2 Admixture Calculator
    By mephisto in forum Inquiries Corner
    Replies: 0
    Last Post: 01-02-2018, 03:48 PM
  5. Maternal Grand Aunt is U4b.
    By jeanL in forum U
    Replies: 4
    Last Post: 01-11-2016, 06:28 AM

Posting Permissions

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