Asking a theoretical question about LDSC

Downstream analysis with genomic data
1

Hi Professors,

I would like to ask a question about LDSC. Since bivariate LDSC could calculate genetic correlation between two traits X and Y, I don’t know whether it is correct to understand the genetic correlation in this way:

for each snp, we could get a correlation between X and Y through their relationships with this snp, and then we could pooled correlations calculated from all snps (just like meta-analysis) and get a pooled correlation which is genetic correlation we get from LDSC?

By the way, I am not a math major, although I understand most of the content in the LDSC video at IBG website, I think this might be more helpful for me to understand the foundation of LDSC in my way if I understant correctly.

Thank you very much and Regards

Xueying

2

Hi Xueying! Basically yes you are correct. The genetic correlation produced by LDSC is simply the correlation of test statistics across all SNPs (Z-scores, or betas) accounting for LD between SNPs.

3

Hi James, thank you for your explanation.

Regards

Xueying

4

Let me know if you need any help running the analysis, feel free to email me at James.Hart@vcuhealth.org
(Also, LDSC is done at the test statisitc level, i.e. Z-score. Not sure why I said beta earlier)