qpDstats produced strange results for some combination of species

Question

qpDstats produced strange results for some combination of species

smallfishcui opened this issue 5 years ago · comments

Hi,

I am using qpdstats in admixtools to detect introgression among species.
I used convertVCFtoEigenstrat.sh script to convert vcf file to eigenstrat file format, and assigned each individual to a population. When I perform the qpDstats, it seems some analyses runs fine, but others just show 0, and the ones run fine were all significant, could anybody tell what's the problem?
Here is part of my result:

W | X | Y | Z | D | stderr | Zscore | BABA | ABBA | nsnps

1	Med	CN	AU	EU	0	1	0	0	0	0
2	Med	CN	AU	Usland	0	1	0	0	0	0
3	Med	CN	AU	Usnat	0	1	0	0	0	0
4	Med	CN	AU	OG	0	1	0	0	0	0
5	Med	CN	EU	AU	0	1	0	0	0	0
6	Med	CN	EU	Usland	0	1	0	0	0	0
7	Med	CN	EU	Usnat	0	1	0	0	0	0
8	Med	CN	EU	OG	0	1	0	0	0	0
9	Med	CN	Usland	AU	0	1	0	0	0	0
10	Med	CN	Usland	EU	0	1	0	0	0	0
37	Med	AU	OG	CN	0	1	0	0	0	0
38	Med	AU	OG	EU	0	1	0	0	0	0
39	Med	AU	OG	Usland	0	1	0	0	0	0
40	Med	AU	OG	Usnat	0	1	0	0	0	0
41	Med	EU	CN	AU	0	1	0	0	0	0
42	Med	EU	CN	Usland	-0,3349	0,035253	-9,5	2485	4986	214014

Thanks,
Cui

bumblenick · Answer 1 · Fri Dec 13 2019 02:58:33 GMT+0800 (China Standard Time)

For all except the last line you have almost no data. Check the AU genotype carefully! Nick

…

On Thu, Dec 12, 2019 at 9:39 AM smallfishcui ***@***.***> wrote: Hi, I am using qpdstats in admixtools to detect introgression among species. I used convertVCFtoEigenstrat.sh script to convert vcf file to eigenstrat file format, and assigned each individual to a population. When I perform the qpDstats, it seems some analyses runs fine, but others just show 0, and the ones run fine were all significant, could anybody tell what's the problem? Here is part of my result: W | X | Y | Z | D | stderr | Zscore | BABA | ABBA | nsnps 37 | Med | AU | OG | CN | 0 | 1 | 0 | 0 | 0 | 0 38 | Med | AU | OG | EU | 0 | 1 | 0 | 0 | 0 | 0 39 | Med | AU | OG | Usland | 0 | 1 | 0 | 0 | 0 | 0 40 | Med | AU | OG | Usnat | 0 | 1 | 0 | 0 | 0 | 0 41 | Med | EU | CN | AU | 0 | 1 | 0 | 0 | 0 | 0 42 | Med | EU | CN | Usland | -0,3349 | 0,035253 | -9,5 | 2485 | 4986 | 214014 Thanks, Cui — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#59?email_source=notifications&email_token=AEE77B5NTIZ2BZI5Q6PTR5TQYJEJPA5CNFSM4JZ7R472YY3PNVWWK3TUL52HS4DFUVEXG43VMWVGG33NNVSW45C7NFSM4IACH6DQ>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AEE77B22QTW7OGJS7VGXMZTQYJEJPANCNFSM4JZ7R47Q> .

CWang_ecoevo · Answer 2 · Fri Dec 13 2019 19:21:10 GMT+0800 (China Standard Time)

Hi Nick, Thank you for the suggestions. Yes, the population AU might be problematic. I guess all of them are somehow problematic, because they are all polyploids and I aligned them against a reference genome of the same species. Now it seems all the individuals have excessive amount of homozygous reference alleles, for example almost all AU sites has the genotype 0/0, and also other populations. In this case I guess even if I filter all the missing data, there are still large amount of SNPs uninformative(I have 776361SNPs in total, missing data filtered out).... So do you have any suggestions for polyploid cases, or is there ways to improve? thanks, Cui bumblenick <notifications@github.com> 于2019年12月12日周四下午8:58写道：

…

For all except the last line you have almost no data. Check the AU genotype carefully! Nick On Thu, Dec 12, 2019 at 9:39 AM smallfishcui ***@***.***> wrote: > Hi, > > I am using qpdstats in admixtools to detect introgression among species. > I used convertVCFtoEigenstrat.sh script to convert vcf file to eigenstrat > file format, and assigned each individual to a population. When I perform > the qpDstats, it seems some analyses runs fine, but others just show 0, and > the ones run fine were all significant, could anybody tell what's the > problem? > Here is part of my result: > > W | X | Y | Z | D | stderr | Zscore | BABA | ABBA | nsnps > 37 | Med | AU | OG | CN | 0 | 1 | 0 | 0 | 0 | 0 > 38 | Med | AU | OG | EU | 0 | 1 | 0 | 0 | 0 | 0 > 39 | Med | AU | OG | Usland | 0 | 1 | 0 | 0 | 0 | 0 > 40 | Med | AU | OG | Usnat | 0 | 1 | 0 | 0 | 0 | 0 > 41 | Med | EU | CN | AU | 0 | 1 | 0 | 0 | 0 | 0 > 42 | Med | EU | CN | Usland | -0,3349 | 0,035253 | -9,5 | 2485 | 4986 | > 214014 > > Thanks, > Cui > > — > You are receiving this because you are subscribed to this thread. > Reply to this email directly, view it on GitHub > < #59?email_source=notifications&email_token=AEE77B5NTIZ2BZI5Q6PTR5TQYJEJPA5CNFSM4JZ7R472YY3PNVWWK3TUL52HS4DFUVEXG43VMWVGG33NNVSW45C7NFSM4IACH6DQ >, > or unsubscribe > < https://github.com/notifications/unsubscribe-auth/AEE77B22QTW7OGJS7VGXMZTQYJEJPANCNFSM4JZ7R47Q > > . > — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#59?email_source=notifications&email_token=AJZNAZNSWWLXTLBTYL7KYIDQYKCVVA5CNFSM4JZ7R472YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEGXVFNA#issuecomment-565138100>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AJZNAZMWRIMF4PBNMDJ3XVLQYKCVVANCNFSM4JZ7R47Q> .

CWang_ecoevo · Answer 3 · Fri Jan 17 2020 00:56:47 GMT+0800 (China Standard Time)

Hi
I come back again.
Still the same problem.
W X Y Z D stderr Zscore BABA ABBA nsnps

1 Med EU Usland CN 0.795 0.140 5.68 196 22 163723
2 Med EU AU CN -0.0341 0.0645 -0.529 16 18 163723
3 Med Usland EU CN 0 1 0 0 0 0
4 Med Usland AU CN 0 1 0 0 0 0
5 Med AU EU CN 0 1 0 0 0 0
6 Med AU Usland CN 0 1 0 0 0 0
7 EU Med Usland CN -0.795 0.140 -5.68 22 196 163723
it seems this is not caused by genotyping problem of a certain population, but for some combinations it go wrong. For example, the data should have 163723SNPs in total, but for those failed lines the SNP is 0. How could it be?

br,
Cui

bumblenick · Answer 4 · Fri Jan 17 2020 04:26:07 GMT+0800 (China Standard Time)

I don't think I can help further. Your data must be weird. Please note that the software assumes biallelic snp data, and I suspect your polyploid data is making it unhappy. Nick

…

On Thu, Jan 16, 2020 at 11:56 AM smallfishcui ***@***.***> wrote: Hi I come back again. Still the same problem. W X Y Z D stderr Zscore BABA ABBA nsnps 1 Med EU Usland CN 0.795 0.140 5.68 196 22 163723 2 Med EU AU CN -0.0341 0.0645 -0.529 16 18 163723 3 Med Usland EU CN 0 1 0 0 0 0 4 Med Usland AU CN 0 1 0 0 0 0 5 Med AU EU CN 0 1 0 0 0 0 6 Med AU Usland CN 0 1 0 0 0 0 7 EU Med Usland CN -0.795 0.140 -5.68 22 196 163723 it seems this is not caused by genotyping problem of a certain population, but for some combinations it go wrong. For example, the data should have 163723SNPs in total, but for those failed lines the SNP is 0. How could it be? br, Cui — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#59?email_source=notifications&email_token=AEE77B4OARNXKWFMXJGIMUDQ6CGU7A5CNFSM4JZ7R472YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEJEY2OY#issuecomment-575245627>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AEE77B2ISTADCTIVQYZMQYTQ6CGU7ANCNFSM4JZ7R47Q> .

CWang_ecoevo · Answer 5 · Fri Jan 17 2020 19:13:31 GMT+0800 (China Standard Time)

Hi Nick,
I just realized one error message from convertf saying that "snp order check fail; snp list not order", as someone poseted here:DReichLab/EIG#37
and here:https://www.biostars.org/p/389958/
So I guess the malformated map file may be the reason that caused the failure of the admixtools run.
Here is the format of my map file produced by vcftools:
7180003098625 340:374:- 0 2078
7180003098625 340:233:- 0 2219
7180003098625 340:213:- 0 2239
7180003098625 340:188:- 0 2264
7180003098625 340:133:- 0 2319
7180003098625 340:83:- 0 2369
According to your instruction I changed the map file to
1 340:374:- 0 2078
1 340:233:- 0 2219
1 340:213:- 0 2239
1 340:188:- 0 2264
1 340:133:- 0 2319
1 340:83:- 0 2369
However, I don't know how to format the $2 and the $4 about the SNP position and coordination, can you give some suggestions?

thanks,
Cui