Summary Stats without EAF & Sample size

[Mahantesh-Biradar]

I have 3 summary stats out of which two of them do not have EAF & sample size columns. I would like to know how I proceed further to use these files as inputs for MTAG. TIA

[paturley]

Hello Mahantesh,

For the allele frequencies, you can just merge your file with the allele frequencies from a reference file of the same ancestries and that should work fine. For the sample size, it's a bit more tricky. If you know the variance of the underlying phenotypes, the effective sample size for each SNP should be

N_eff = sig2_y/(2EAF(1-EAF)*SE^2)

where sig2_y is the phenotypic variance, EAF is the effect allele frequency (from a reference panel if necessary), and SE is the reported standard error. I've had some mixed success with this though. For example, if your SEs are rounded to only a few decimals, then N_eff can be a bit wonky.

Hope this helps.

Best, Patrick

On Tue, Mar 15, 2022 at 9:53 AM Mahantesh Biradar @.***> wrote:

I have 3 summary stats out of which two of them do not have EAF & sample size columns. I would like to know how I proceed further to use these files as inputs for MTAG. TIA

— Reply to this email directly, view it on GitHub https://github.com/JonJala/mtag/issues/154, or unsubscribe https://github.com/notifications/unsubscribe-auth/AFBUB5OIZVJ5FFNO24A6S7TVACI4VANCNFSM5QYYCPKQ . Triage notifications on the go with GitHub Mobile for iOS https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675 or Android https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub.

You are receiving this because you are subscribed to this thread.Message ID: @.***>

[Mahantesh-Biradar]

Hi Patrick, thanks for your explanation. What will be a formula to calculate N_eff if my phenotype is binary?

[paturley]

Was the GWAS done with a linear probability model or logistic regression?

On Tue, Mar 15, 2022 at 11:22 AM Mahantesh Biradar @.***> wrote:

Hi Patrick, thanks for your explanation. What will be a formula to calculate N_eff if my phenotype is binary?

— Reply to this email directly, view it on GitHub https://github.com/JonJala/mtag/issues/154#issuecomment-1068116037, or unsubscribe https://github.com/notifications/unsubscribe-auth/AFBUB5IVT57CYOQQ7NVA4JLVACTMHANCNFSM5QYYCPKQ . Triage notifications on the go with GitHub Mobile for iOS https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675 or Android https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub.

You are receiving this because you commented.Message ID: @.***>

[Mahantesh-Biradar]

It should be logistic regression

[paturley]

For logistic regression, check out issue #60 . There's a standard formula for effective N in that case.

[Mahantesh-Biradar]

Thank you very much Patrick!

[Mahantesh-Biradar]

Hi Patrick, Can you please suggest what is the best way to obtain 'sig2_y' for a summary statistic of quantitative trait?

[paturley]

If you don't have the data, then hopefully they report it in the paper. In many cases, the phenotype has been standardized, which would mean sig2_y=1.

On Tue, Mar 22, 2022 at 7:34 AM Mahantesh Biradar @.***> wrote:

Hi Patrick, Can you please suggest what is the best way to obtain 'sig2_y' for a summary statistic of quantitative trait?

— Reply to this email directly, view it on GitHub https://github.com/JonJala/mtag/issues/154#issuecomment-1075066549, or unsubscribe https://github.com/notifications/unsubscribe-auth/AFBUB5J6VCNWFLEAHHAZHRTVBGV27ANCNFSM5QYYCPKQ . You are receiving this because you commented.Message ID: @.***>

[Mahantesh-Biradar]

Got it, Paul. Thanks for the clarification.