Open anbai106 opened 2 years ago
I am wondering if there are people maintaining this wonder software :)
I've been encountering this same issue recently (large rG values but wide SEs). The 'problem' is that if either or both traits have low heritability then the rG estimate is inherently unstable. In your case the H2 estimate for brain volume is not significantly different from zero (h2=0.0144, se=0.026, therefore p=0.58). It would be helpful if LDSC flagged this in the output more explicitly. In the munge step you get a warning that mean chi^2 may be too small but easy to miss. Does seem surprising it would not be significantly different to zero (quantitative trait in 18,000 people?). Not SNPs p<5*10-8 though?
Hi,
Thanks for this wonderful tool!
I am running a genetic correlation between brain volume (C256_55) and AD. I obtained very low h2 estimates for this brain volume (I ran GWAS with) and also AD (downloaded from GWAS Catalog). I also ran GCTA to estimate h2 for the brain volume, and I obtained: 0.135862 +- 0.026911.
Here is the output from ldsc:
This is the output from the munge_stat step:
More information for the AD GWAS summary, I downloaded it from this here: http://ftp.ebi.ac.uk/pub/databases/gwas/summary_statistics/GCST002001-GCST003000/GCST002245/
The log for the mung_stat:
Then I estimated the h2 for AD trait:
Any insights into these results? My sample size is large enough: N > 18K