Post by elias on Mar 9, 2016 5:08:57 GMT
Hi!
I recently was able to run the reml LD-MS analysis on GTCA.
I run a control trait (BMI) to check that my estimate is consistent with previous study and it worked fine:
(BMI)
--mgrm-gz list.of.GRMs.txt
--pheno GCTA_phenotype.BMI.txt
--reml
--out BMI.Estimate.Imp.GCTA.GRM
--thread-num 6
Note: the program will be running on 6 threads.
Reading phenotypes from [GCTA_phenotype.BMI.txt].
Non-missing phenotypes of 27556 individuals are included from [GCTA_phenotype.BMI.txt].
There are 28 GRM file names specified in the file [list.of.GRMs.txt].
Sum of V(G)/Vp 0.349805 0.019091
Then I run the analysis on my trait of interest but the results is not consistent with the estimate I got when running one single GRM (non LD-MS):
*******************************************************************
* Genome-wide Complex Trait Analysis (GCTA)
* version 1.25.2
* (C) 2010-2013 Jian Yang, Hong Lee, Michael Goddard and Peter Visscher
* The University of Queensland
* MIT License
*******************************************************************
Analysis started: Mon Mar 7 15:41:02 2016
Options:
--mgrm-gz list.of.GRMs.txt
--covar GCTA_phenotype.BCOVAR1.txt
--prevalence 0.13
--qcovar GCTA_phenotype.QCOVAR1.txt
--pheno GCTA_phenotype.txt
--reml
--out Vasc_Adjs.Estimate.Imp.GCTA.GRM
--thread-num 6
Note: the program will be running on 6 threads.
Sum of V(G)/Vp 0.464961 0.021789
The estimate of variance explained on the observed scale is transformed to that on the underlying scale:
(Proportion of cases in the sample = 0.131461; User-specified disease prevalence = 0.130000)
Sum of V(G)_L/Vp 1.163990 0.054547
This estimate is very high and not consistent compare to a single GRM using the same phenotype and co-variate:
Analysis started: Tue Jan 19 14:15:50 2016
Options:
--grm-gz GCTA-GWAS.1000G
--covar GCTA_phenotype.BCOVAR1.txt
--prevalence 0.13
--qcovar GCTA_phenotype.QCOVAR1.txt
--pheno GCTA_phenotype.txt
--reml
--out Vasc_Adjs.Estimate.Imp.GCTA.GRM
--thread-num 6
Note: the program will be running on 6 threads.
V(G)/Vp 0.107266 0.012601
The estimate of variance explained on the observed scale is transformed to that on the underlying scale:
(Proportion of cases in the sample = 0.131461; User-specified disease prevalence = 0.130000)
V(G)/Vp_L 0.268530 0.031545
I would really appreciate your suggestion on this analysis and your explanation on why estimate of unrelated samples is over 1.
Thanks!,
Elias