PDA

View Full Version : Deviation of G25 coords simulated by K36 data



rothaer
12-28-2022, 11:51 PM
This is the deviation of G25 coords - simulated by K36 data - from the real G25 coords as it shows in the North Europe PCA. I depicted the deviations by a red arrow going to the simulated coords’ dot. Just in five cases where the deviation was very small I did not draw an arrow (Drewcastle, Feiichy, Freiichy_father, ph2ter_wife and IrisSelene).

https://i.imgur.com/xk7mZSr.jpg

Kiel
12-29-2022, 04:51 PM
heres mine.
https://i.ibb.co/WvRWtLG/wog.png (https://ibb.co/JjSRQ9m)

Luke35
12-29-2022, 05:16 PM
Nice visualization, rothaer.

I've been saying for a while that I think it is important to simulate your source populations and your target populations all from K36, or all from K13, whatever you are using just keep it the same. From what I have seen, that produces a reasonable result.

I think the accuracy gets notably worse when you combine real G25 coordinates with simulated ones.

cakmir7y
12-29-2022, 05:57 PM
Tried mine

https://i.imgur.com/2Xdzcs3.png

rothaer
12-29-2022, 06:33 PM
Nice visualization, rothaer.

I've been saying for a while that I think it is important to simulate your source populations and your target populations all from K36, or all from K13, whatever you are using just keep it the same. From what I have seen, that produces a reasonable result.

I think the accuracy gets notably worse when you combine real G25 coordinates with simulated ones.

Exactly my thoughts.

The main problem is that there are no simulated by K36 data G25 coords for ancient DNA. And, btw. if you would have the K36 data for all you could go directly by K36 instead of G25.

Now, somewhere Davidski stated that K36 is designed that way that it catches modern genetic drifts, while G25 is not and hence more useful for ancient DNA. If this is true, G25 should be more suitable when we deal with old references. Which in turn would suggest real G25 coords...

Luke35
12-29-2022, 06:41 PM
Exactly my thoughts.

The main problem is that there are no simulated by K36 data G25 coords for ancient DNA. And, btw. if you would have the K36 data for all you could go directly by K36 instead of G25.

Now, somewhere Davidski stated that K36 is designed that way that it catches modern genetic drifts, while G25 is not and hence more useful for ancient DNA. If this is true, G25 should be more suitable when we deal with old references. Which in turn would suggest real G25 coords...

Absolutely, I would just use K36 in the case of having that information/data, no sense in converting other than for fun (for use with moderns).

I do think converting K13 to G25 is interesting because the higher number of coordinates (in G25) might theoretically allow for more nuance/subtlety in modeling.

Kiel
12-29-2022, 06:55 PM
Just wanted to share this. K15 and K13 are way closer for me, especially K15. Maybe the best way to sim g25 would be to run all 3 and make averages (if its possible to make averages out of those?)
https://i.ibb.co/h8ZdVh0/sim.png (https://ibb.co/HgxPq1Z)