search

DReichLab / EIG

Eigen tools by Nick Patterson and Alkes Price lab
Other
180 stars 60 forks source link

Eigenstrat for 3 subpopulations #66

Open tahseen-igib opened 3 years ago

tahseen-igib commented 3 years ago

Hi I have a list of SNPs from 3 subpopulations pairwise association tests (1vs2 ,2vs3, 1vs3) that i need to check for population stratification.I have PC1-10 values for all the 3 subpopulations taken together. However,in eigenstrat we have to define case-control to samples for chi-square test which in my case i have only for 2 subpopulations at a time.

Is it technically correct to take PC1-10 for 2 subgroups at a time and do the correction using eigenstrat? Kindly suggest

bumblenick commented 3 years ago

I'm not sure about your needs but if you want to understand the structure smartpca 1) Will tell you about population differentiation along PCA axes 2) Calculates F_st If you are trying to do a GWAS eigenstrat seems to already meet your needs.

I suspect I need a clear description about what you are trying to do.

Nick

On Sun, Jun 6, 2021 at 7:01 AM tahseen-igib @.***> wrote:

Hi I have a list of SNPs from 3 subpopulations pairwise association tests (1vs2 ,2vs3, 1vs3) that i need to check for population stratification.I have PC1-10 values for all the 3 subpopulations taken together. However,in eigenstrat we have to define case-control to samples for chi-square test which in my case i have only for 2 subpopulations at a time.

Is it technically correct to take PC1-10 for 2 subgroups at a time and do the correction using eigenstrat? Kindly suggest

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/DReichLab/EIG/issues/66, or unsubscribe https://github.com/notifications/unsubscribe-auth/AEE77B3RF465WAYUWDA6Z7DTRNIOZANCNFSM46FZVYZQ .

tahseen-igib commented 3 years ago

Hi Nick

I have a deep phenotyped exome data that is subgrouped into 3 subpopulations based on some phenotypes.I am trying to understand genetic differences between those groups. Initially i have done PCA (using smartPCA) for complete population (all 3 groups included). I have done aasociation test (fisher's exact test) by taking pair of groups at a time (1vs2,2vs3,1vs3).Each pairwise comparison has given me a list of significant SNPs.I want to filter out SNPs that have come due to population stratification and are not true signals. For this i looked for Eigenstrat that will give chi-square corrected p value (correction for stratification by providing it PC components).& this needs our samples to be defined as case-control. But in my case, i have got SNPs from paired group at a time & my PCs are representative of complete population (all 3 groups), my query is while doing eigenstrat for 1vs2 group SNPs (my case control will be 1 & 2 group) can i eliminate PCs from 3rd group??

I hope i am able to give u clear description of my problem

Thanks

bumblenick commented 3 years ago

Please send me email. This thread is not of general interest.

N

On Sun, Jun 6, 2021 at 11:17 AM tahseen-igib @.***> wrote:

Hi Nick

I have a deep phenotyped exome data that is subgrouped into 3 subpopulations based on some phenotypes.I am trying to understand genetic differences between those groups. Initially i have done PCA (using smartPCA) for complete population (all 3 groups included). I have done aasociation test (fisher's exact test) by taking pair of groups at a time (1vs2,2vs3,1vs3).Each pairwise comparison has given me a list of significant SNPs.I want to filter out SNPs that have come due to population stratification and are not true signals. For this i looked for Eigenstrat that will give chi-square corrected p value (correction for stratification by providing it PC components).& this needs our samples to be defined as case-control. But in my case, i have got SNPs from paired group at a time & my PCs are representative of complete population (all 3 groups), my query is while doing eigenstrat for 1vs2 group SNPs (my case control will be 1 & 2 group) can i eliminate PCs from 3rd group??

I hope i am able to give u clear description of my problem

Thanks

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/DReichLab/EIG/issues/66#issuecomment-855415327, or unsubscribe https://github.com/notifications/unsubscribe-auth/AEE77B6G643Y5O4UZBL3JMDTROGR5ANCNFSM46FZVYZQ .

tahseen-igib commented 3 years ago

Sure.I have sent u mail at nickp@broadinstitute.org.