Is "Lambda GC" valid when h2 is very low?

[Jiang]

Hi LDSC experts,

I ran a GWAS in the UK Biobank --with relatively few cases--, and I'm getting a very low h2 estimate, similar to what is described here. More specifically, I have these two scenarios (two munge stats and two h2 files):

GWAS #1 Munge stats file:

Mean chi^2 = 0.999
WARNING: mean chi^2 may be too small.
Lambda GC = 1.001
Max chi^2 = 25.326
0 Genome-wide significant SNPs (some may have been removed by filtering).

GWAS #1 h2 file:

Total Observed scale h2: 0.0015 (0.0015)
Lambda GC: 1.0075
Mean Chi^2: 1.003
Intercept: 0.9943 (0.0069)
Ratio < 0 (usually indicates GC correction).

GWAS #2 Munge stats file:

Mean chi^2 = 0.999
WARNING: mean chi^2 may be too small.
Lambda GC = 0.997
Max chi^2 = 23.817

GWAS #2 h2 file:

Total Observed scale h2: 0.0002 (0.0015)
Lambda GC: 1.0016
Mean Chi^2: 1.0017
Intercept: 1.0007 (0.0067)
Ratio: 0.3995 (3.9369)

In GWAS #2, I'm actually getting "NA"s when I run genetic correlation analysis (via LD Hub), such as in the posts here and here, which is apparently due to the very low h2.

Long story short, I have a question related to this: if I do not care about heritability, may I use the "Lambda GC" values shown above as an evidence of lack of inflation in the GWAS? If so, which "Lambda GC" value is best (from munge stats file or from h2 file)?


Thanks!

P.S.: By the way, should I still *trust* the genetic correlation results of GWAS #1 versus multiple other traits (things run smoothly on LD Hub)?

[Raymond Walters]

Hi,

Yes, your lambdaGC values would support the conclusion that there little or no inflation in your GWAS. 

If you are just looking for lambdaGC for this purpose, then it’s probably better to compute it yourself on your full results rather than use these reported values after ldsc’s additional filtering. If you’re looking for ldsc’s inflation measure, that would be the intercept reported in the h2 logs.

I’d be very skeptical of genetic correlation results for your GWAS #1. Generally we don’t recommend running the rg analysis  with traits that don’t have significant h2 results. (I’m actually a bit surprised you aren’t getting NAs with GWAS #1 also.)

Cheers,

Raymond

--
You received this message because you are subscribed to the Google Groups "ldsc_users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ldsc_users+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/ldsc_users/bc211ead-7fb7-4b4e-a36e-16dcff2ed88b%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.