The huge discrepancy between z_mtag and z_gwas

[kys21207]

I expected that Z_mtag and Z_gwas are near perfect fit like the mtag paper. However, my results showed that there is a huge discrepancy between z_mtag and z_gwas as below. (R^2=0.41 and slope=0.6) Do you think it's likely caused by the violation of mtag's assumptions? The plot was drawn based on trait2.

I ran mtag with only two traits. Estimated Omega: [[9.676e-07 1.276e-06] [1.276e-06 4.481e-06]]

(Correlation): [[1. 0.613] [0.613 1. ]]

Estimated Sigma: [[0.928 0.114] [0.114 0.984]]

(Correlation): [[1. 0.119] [0.119 1. ]]

mean of chi^2: trait1=1.06, trait2=1.054 MTAG Max FDR: trait1=0.0009, trait2=0.14 GWAS Max FDR: trait1=0.0349, trait2=0.10

[paturley]

In the MTAG paper, the MTAG and GWAS z-stats were only similar when the MTAG was based on a subsets of the UKB and the GWAS was based on the full UKB sample. It was meant as an illustration that MTAG correctly accounts for overlap. In your case, it looks like you are comparing the z-stats of the input GWAS for one trait to the z-stats of the output MTAG combining two traits. If the MTAG z-stats were identical to the input GWAS z-stats, that would imply that there was no gain in power from MTAG, which would make MTAG pretty useless. Looking at all the information you posted here, it all looks pretty sensible to me.

Let me know if I misunderstood what you did though.

On Fri, Oct 16, 2020 at 9:37 AM Kijoung Song [email protected] wrote:

I expected that Z_mtag and Z_gwas are near perfect fit like the mtag paper. However, my results showed that there is a huge discrepancy between z_mtag and z_gwas as below. (R^2=0.41 and slope=0.6) [image: gwas_mtag_zscore_comp] https://user-images.githubusercontent.com/20864872/96263918-99351700-0f91-11eb-93a1-1fb6ccddb25b.png Do you think it's likely caused by the violation of mtag's assumptions? The plot was drawn based on trait2.

I ran mtag with only two traits. Estimated Omega: [[9.676e-07 1.276e-06] [1.276e-06 4.481e-06]]

(Correlation): [[1. 0.613] [0.613 1. ]]

Estimated Sigma: [[0.928 0.114] [0.114 0.984]]

(Correlation): [[1. 0.119] [0.119 1. ]]

mean of chi^2: trait1=1.06, trait2=1.054 MTAG Max FDR: trait1=0.0009, trait2=0.14 GWAS Max FDR: trait1=0.0349, trait2=0.10

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/JonJala/mtag/issues/115, or unsubscribe https://github.com/notifications/unsubscribe-auth/AFBUB5PHU2VCMUPPRUECF3DSLBEALANCNFSM4STLJDVQ .

[kys21207]

Thank you. It makes sense.