search

DReichLab / AdmixTools

Tools test whether admixture occurred and more
183 stars 64 forks source link

F3 test detect somewhat ancient gene flow #20

Open biozzq opened 7 years ago

biozzq commented 7 years ago

Dear @bumblenick

I have both wild and domesticated populations. From the analysis of inferred ancestral components (using admixture software), I found some wild individuals contained domesticate components (about 50%). So I want to use the three-population test to test whether these individuals have gene flow from the domesticated populations. But the results are all non-negative, meanwhile, I also do this at the individual level, the results are same as the population level. But from the results, I can see the f_3 values for these admixtured individuals (SRR1576679,BS1,ERR340348) are smaller than the unadmixtured individuals (ERR340329,ERR340333,ERR340335,ERR340341) when do analysis at individual level. I guess this admixture event may be ancient (about several thousands generations ago). So I want to know can I get some more information from the following results when the f_3 value is non-negative and can three-population test detect ancient gene flow? Thanks.

snps: 16016002 Source 1 Source 2 Target f_3 std. err Z SNPs result: ERR340334 A14 BS1 0.079671 0.005448 14.624 12914067 result: ERR340334 A14 ERR340329 0.602950 0.036776 16.395 13012058 result: ERR340334 A14 ERR340333 0.404647 0.025341 15.968 13052431 result: ERR340334 A14 ERR340335 0.334453 0.020576 16.255 13053033 result: ERR340334 A14 ERR340341 0.487972 0.026363 18.510 12994993 result: ERR340334 A14 ERR340348 0.106602 0.006818 15.636 12944510 result: ERR340334 A14 SRR1576679 0.084376 0.004683 18.016 12857869

Best Zhuqing

bumblenick commented 7 years ago

In general a positive score fails to prove gene flow. You may be able to use f4 (qpDstat) to at least show that a simple tree model fails.

Nick

On Sun, Mar 26, 2017 at 10:49 AM, biozzq notifications@github.com wrote:

Dear @bumblenick https://github.com/bumblenick

I have both wild and domesticated populations. From the analysis of inferred ancestral components (using admixture software), I found some wild individuals contained domesticate components (about 50%). So I want to use the three-population test to test whether these individuals have gene flow from the domesticated populations. But the results are all non-negative, meanwhile, I also do this at the individual level, the results are same as the population level. But from the results, I can see the f_3 values for these admixtured individuals (SRR1576679,BS1,ERR340348) are smaller than the unadmixtured individuals (ERR340329,ERR340333,ERR340335,ERR340341) when do analysis at individual level. I guess this admixture event may be ancient (about several thousands generations ago). So I want to know can I get some more information from the following results when the f_3 value is non-negative and can three-population test detect ancient gene flow? Thanks.

snps: 16016002 Source 1 Source 2 Target f_3 std. err Z SNPs result: ERR340334 A14 BS1 0.079671 0.005448 14.624 12914067 result: ERR340334 A14 ERR340329 0.602950 0.036776 16.395 13012058 result: ERR340334 A14 ERR340333 0.404647 0.025341 15.968 13052431 result: ERR340334 A14 ERR340335 0.334453 0.020576 16.255 13053033 result: ERR340334 A14 ERR340341 0.487972 0.026363 18.510 12994993 result: ERR340334 A14 ERR340348 0.106602 0.006818 15.636 12944510 result: ERR340334 A14 SRR1576679 0.084376 0.004683 18.016 12857869

Best Zhuqing

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/DReichLab/AdmixTools/issues/20, or mute the thread https://github.com/notifications/unsubscribe-auth/AQn_h8xSm0yYlPQmoxKhzuWjd0UUMBPZks5rpnsBgaJpZM4Mpcxp .

biozzq commented 7 years ago

Dear @bumblenick

Thank you. Can I do f3 test and qpDstat at individual level? What is more, You mean that a positive score is meaningless on gene flow. When the score is negative, it can prove the gene flow from both source individuals or populations. If the gene flow just from one source population, will the score be negative?

Best Zhuqing

biozzq commented 7 years ago

Dear @bumblenick

I found I can use F4 ratio test to estimate the admixture proportions and the result seems appropriate. So I want to know whether it is normal that three-population test failed but succeeded in F4-ratio estimation. More, I do not have an accurate genetic map which just set from physical distance, does genetic map effect the results, like f3-test, D-statistic and F4-ratio test? Many thanks

Best Zhuqing

bumblenick commented 7 years ago

1) f3 is sensitive to post admixture drift, but not f4. Read my 2012 paper if you can. 2) On genetic map, it's OK if you have physical map, and LD does not extend much beyond 5M bases.

N

On Mon, Mar 27, 2017 at 12:26 PM, biozzq notifications@github.com wrote:

Dear @bumblenick https://github.com/bumblenick

I found I can use F4 ratio test to estimate the admixture proportions and the result seems appropriate. So I want to know whether it is normal that three-population test failed but succeeded in F4-ratio estimation. More, I do not have an accurate genetic map which just set from physical distance, does genetic map effect the results, like f3-test, D-statistic and F4-ratio test? Many thanks

Best Zhuqing

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/DReichLab/AdmixTools/issues/20#issuecomment-289507079, or mute the thread https://github.com/notifications/unsubscribe-auth/AQn_h8rHUzEhUwCGaNN-Jj-PY6yJuAjvks5rp-NTgaJpZM4Mpcxp .

biozzq commented 7 years ago

Dear @bumblenick

Thanks. Do you mean if the target population experienced a high degree of specific genetic drift after admixture, f3 will fail to be negative?

More, when I do f4, I do not know how to chose a right combination between A and B. Should I assure the population A have no genetic contribute to B? If A have admixture with B, does it mean there are many possible for the admixture event 1, the admixture event occurred in the most common ancestral of A and B 2, their admixture event occurred after splitting A and B, but the admixture source may be difficult to distinguish A from B.

Best Wishes Zhuqing

bumblenick commented 7 years ago

1) Read the 2012 ancient admixture paper as carefully as you can! f3 can indeed fail to be negative if there is a lot of drift in target

2) From f4 you can often tell that a simple tree is wrong, but then you have to think! Joe Pickrel's treemix is something that you might want to experiment with.

Nick

On Fri, Mar 31, 2017 at 10:45 PM, biozzq notifications@github.com wrote:

Dear @bumblenick https://github.com/bumblenick

Thanks. Do you mean if the target population experienced a high degree of specific genetic drift after admixture, f3 will fail to be negative?

More, when I do f4, I do not know how to chose a right combination between A and B. Should I assure the population A have no genetic contribute to B? If A have admixture with B, does it mean there are many possible for the admixture event 1, the admixture event occurred in the most common ancestral of A and B 2, their admixture event occurred after splitting A and B, but the admixture source may be difficult to distinguish A from B.

Best Wishes Zhuqing

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/DReichLab/AdmixTools/issues/20#issuecomment-290888704, or mute the thread https://github.com/notifications/unsubscribe-auth/AQn_h8or2zfxfSW2uqK9Ka4mxdM_TRgLks5rrboxgaJpZM4Mpcxp .