paturley
commented
1 year ago So it seems like your setting is the highest risk scenario. The pre-MTAG mean chi2 statistic for trait 1 is 1.1, but it is 1.6 for trait two, meaning that the GWAS for trait 2 is around 6 times better powered than the one for trait one. On top of that, the genetic correlation of the traits seem to be about .6. As we describe in the paper, maxFDR is lowest when there are large power differentials between pairs of traits and a moderate genetic correlation.
That doesn't mean that the FDR is .4, but it does mean that it could be that high if the genetic architecture is conspiring against you. I don't think there is a way to "fix" it other than just reporting the number and saying that, as a worst case scenario, 40% of the genome-wide significant hits may be false positives.
On Mon, Jul 31, 2023 at 4:28 AM Donald Sandoz @.***> wrote:
Summary of MTAG results:
Trait # SNPs used N (max) N (mean) GWAS mean chi^2 MTAG mean chi^2 GWAS equiv. (max) N 1 ...peline/QC/trait1.txt 5820480 964057 950191 1.132 1.340 2476641 2 ...eline/QC/trait2.txt 5820480 1165720 1147727 1.620 1.634 1191479 3 ...peline/QC/trait3.txt 5820480 1030836 1030836 1.364 1.371 1049044
Estimated Omega: [[1.396e-07 1.806e-07 1.223e-07] [1.806e-07 5.429e-07 8.903e-08] [1.223e-07 8.903e-08 3.716e-07]]
(Correlation): [[1. 0.656 0.537] [0.656 1. 0.198] [0.537 0.198 1. ]]
Estimated Sigma: [[1.007 0.152 0.246] [0.152 1.012 0.113] [0.246 0.113 1.051]]
(Correlation): [[1. 0.151 0.239] [0.151 1. 0.11 ] [0.239 0.11 1. ]]
MTAG weight factors: (average across SNPs) [0.907 1.085 1.105]
2023/07/30/03:07:43 PM 2023/07/30/03:07:43 PM MTAG results saved to file. 2023/07/30/03:07:43 PM Beginning maxFDR calculations. Depending on the number of grid points specified, this might take some time... 2023/07/30/03:07:43 PM T=3 2023/07/30/03:07:46 PM Number of gridpoints to search: 1554 2023/07/30/03:07:46 PM Performing grid search using 1 cores. 2023/07/30/03:07:46 PM Grid search: 10.0 percent finished for . Time: 0.009 min 2023/07/30/03:07:47 PM Grid search: 20.0 percent finished for . Time: 0.019 min 2023/07/30/03:07:47 PM Grid search: 30.0 percent finished for . Time: 0.028 min 2023/07/30/03:07:48 PM Grid search: 40.0 percent finished for . Time: 0.037 min 2023/07/30/03:07:49 PM Grid search: 50.0 percent finished for . Time: 0.047 min 2023/07/30/03:07:49 PM Grid search: 60.0 percent finished for . Time: 0.056 min 2023/07/30/03:07:50 PM Grid search: 70.0 percent finished for . Time: 0.066 min 2023/07/30/03:07:50 PM Grid search: 80.0 percent finished for . Time: 0.076 min 2023/07/30/03:07:51 PM Grid search: 90.0 percent finished for . Time: 0.085 min 2023/07/30/03:07:51 PM Grid search: 100.0 percent finished for . Time: 0.095 min 2023/07/30/03:07:52 PM Saved calculations of fdr over grid points in E:/OneDrive/pipeline/MTAG/fdr_mat.txt 2023/07/30/03:07:52 PM <><><<>><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><> 2023/07/30/03:07:52 PM grid point indices for max FDR for each trait: [1045 82 35] 2023/07/30/03:07:52 PM Maximum FDR 2023/07/30/03:07:52 PM Max FDR of Trait 1: 0.413430258303 at probs = [0.2
- 0.1 0. 0.3 0.4 0. ] 2023/07/30/03:07:52 PM Max FDR of Trait 2: 0.000286570446712 at probs = [0. 0. 0. 0. 0.3 0. 0. 0.7] 2023/07/30/03:07:52 PM Max FDR of Trait 3: 0.00224058220201 at probs = [0.
- 0.5 0.3 0.2] 2023/07/30/03:07:52 PM <><><<>><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><> 2023/07/30/03:07:52 PM Completed FDR calculations. 2023/07/30/03:07:52 PM MTAG complete. Time elapsed: 15.0m:42.7620000839s Above is the maxFDR calculated from the MTAG analysis I ran for the three traits, with a surprisingly high value of 0.4 for trait 1. I then ran the MTAG analyses two by two, and the maxFDR for trait 1 was 0.03, and the max-FDR I got from running the MTAG analyses for that one trait alone was 0.03. Here I found very little change in mean chi^2, only that GWAS equiv. (max) N changed a lot could that be the problem? How it is calculated can you tell me? Is there any way I can solve this problem or my data is not applicable to this method? By the way, the GWAS equiv. (max) N has grown to that number when I performed the MTAG analysis for trait 1 and trait 2, not for trait 1 and trait 3.
— Reply to this email directly, view it on GitHub https://github.com/JonJala/mtag/issues/182, or unsubscribe https://github.com/notifications/unsubscribe-auth/AFBUB5OH5XWMHTCG5W73QQTXS5UCDANCNFSM6AAAAAA254ZBFY . You are receiving this because you are subscribed to this thread.Message ID: @.***>
DonaldSandoz2000
Summary of MTAG results:
Trait # SNPs used N (max) N (mean) GWAS mean chi^2 MTAG mean chi^2 GWAS equiv. (max) N 1 ...peline/QC/trait1.txt 5820480 964057 950191 1.132 1.340 2476641
2 ...eline/QC/trait2.txt 5820480 1165720 1147727 1.620 1.634 1191479
3 ...peline/QC/trait3.txt 5820480 1030836 1030836 1.364 1.371 1049044
Estimated Omega: [[1.396e-07 1.806e-07 1.223e-07] [1.806e-07 5.429e-07 8.903e-08] [1.223e-07 8.903e-08 3.716e-07]]
(Correlation): [[1. 0.656 0.537] [0.656 1. 0.198] [0.537 0.198 1. ]]
Estimated Sigma: [[1.007 0.152 0.246] [0.152 1.012 0.113] [0.246 0.113 1.051]]
(Correlation): [[1. 0.151 0.239] [0.151 1. 0.11 ] [0.239 0.11 1. ]]
MTAG weight factors: (average across SNPs) [0.907 1.085 1.105]
2023/07/30/03:07:43 PM
2023/07/30/03:07:43 PM MTAG results saved to file. 2023/07/30/03:07:43 PM Beginning maxFDR calculations. Depending on the number of grid points specified, this might take some time... 2023/07/30/03:07:43 PM T=3 2023/07/30/03:07:46 PM Number of gridpoints to search: 1554 2023/07/30/03:07:46 PM Performing grid search using 1 cores. 2023/07/30/03:07:46 PM Grid search: 10.0 percent finished for . Time: 0.009 min 2023/07/30/03:07:47 PM Grid search: 20.0 percent finished for . Time: 0.019 min 2023/07/30/03:07:47 PM Grid search: 30.0 percent finished for . Time: 0.028 min 2023/07/30/03:07:48 PM Grid search: 40.0 percent finished for . Time: 0.037 min 2023/07/30/03:07:49 PM Grid search: 50.0 percent finished for . Time: 0.047 min 2023/07/30/03:07:49 PM Grid search: 60.0 percent finished for . Time: 0.056 min 2023/07/30/03:07:50 PM Grid search: 70.0 percent finished for . Time: 0.066 min 2023/07/30/03:07:50 PM Grid search: 80.0 percent finished for . Time: 0.076 min 2023/07/30/03:07:51 PM Grid search: 90.0 percent finished for . Time: 0.085 min 2023/07/30/03:07:51 PM Grid search: 100.0 percent finished for . Time: 0.095 min 2023/07/30/03:07:52 PM Saved calculations of fdr over grid points in E:/OneDrive/pipeline/MTAG/fdr_mat.txt 2023/07/30/03:07:52 PM <><><<>><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><> 2023/07/30/03:07:52 PM grid point indices for max FDR for each trait: [1045 82 35] 2023/07/30/03:07:52 PM Maximum FDR 2023/07/30/03:07:52 PM Max FDR of Trait 1: 0.413430258303 at probs = [0.2 0. 0. 0.1 0. 0.3 0.4 0. ] 2023/07/30/03:07:52 PM Max FDR of Trait 2: 0.000286570446712 at probs = [0. 0. 0. 0. 0.3 0. 0. 0.7] 2023/07/30/03:07:52 PM Max FDR of Trait 3: 0.00224058220201 at probs = [0. 0. 0. 0. 0. 0.5 0.3 0.2] 2023/07/30/03:07:52 PM <><><<>><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><> 2023/07/30/03:07:52 PM Completed FDR calculations. 2023/07/30/03:07:52 PM MTAG complete. Time elapsed: 15.0m:42.7620000839s Above is the maxFDR calculated from the MTAG analysis I ran for the three traits, with a surprisingly high value of 0.4 for trait 1. I then ran the MTAG analyses two by two, and the maxFDR for trait 1 was 0.03, and the max-FDR I got from running the MTAG analyses for that one trait alone was 0.03. Here I found very little change in mean chi^2, only that GWAS equiv. (max) N changed a lot could that be the problem? How it is calculated can you tell me? Is there any way I can solve this problem or my data is not applicable to this method? By the way, the GWAS equiv. (max) N has grown to that number when I performed the MTAG analysis for trait 1 and trait 2, not for trait 1 and trait 3.