Low heritabiltiy and genetic correlations

[demodw]

Hi,

I tried to use HDL to get a more accurate estimate of rg for two phenotype with low heritability. in LDSC the heritabilities are estimated to be 0.0077 (0.0017) and 0.0054 (0.0018). The genetic correlation is -0.18. However, when I run HDL, the heritabilities are estimated to be 0.0044 (0.0013) and 0 (0.002). It's two binary phenotypes, and the sample size is in both cases >248k (N cases + N controls). For both phenotypes, >99% of the SNPs are in the reference. It's a Western European population. I've added a screenshot of the HDL and LDSC output.

What I'm curious about is why the heritability for phenotype 2 is 0, thus making it not possible to estimate rg. Is the heritability of phenotype 2 simply too low, despite being only ~30% lower than phentoype 1, with the ~same standard error? Is there something else I can do to calculate the rg using HDL?

Edit: Looking into the code of HDL, I see that the beta values are approximated using Z / sqrt(N). However, my summary stats includes relateds and was estimated using mixed models. Thus, the N is not entirely correct, I guess. When I compare the bhat.raw versus the observed ones, the bhat.raw is ~10x lower than the estimated one. Could this be an issue?

Thanks in advance.