Ancestry help

[Lora Pless]

Hello, I'm interested to run ldsc in the R package to estimate heritability for two phenotypes,  as well as the genetic correlation estimate. I've successfully completed this using GWAS summary stats for individuals of European ancestry. I'd next like to do the same for individuals of African ancestry. Are the following files, as required to run munge and ldsc, available for individuals of African ancestry?

Files:

1. Hm3 file
2. ld
3. wld

Thank you in advance!

[agro...@gmail.com]

Thanks for asking this question. The HM3 file is just used for aligning to the same reference allele across summary statistics and could, in theory, still be used for African ancestry groups (though you would potentially want to update this depending on whether you have a very different set of variants in your ld and wld files). 

The ld and wld file are often specified to be the same. African ancestry LD-scores are provided on the panUKB website (I can't remember if there is any reformatting needed to get those to work but the website indicates they are LDSC compatible):

https://pan.ukbb.broadinstitute.org/downloads

Another option is to create your own LD-scores using public data resources. For example, 1000G Phase3 data contains both (i) an African Ancestry in the southwest United States subgroup (n = 61); (ii) and an African population (n = 661) group. This is quite a bit more work, but details on how to create your own LD-scores are provided by the original developers here: 

https://github.com/bulik/ldsc/wiki/LD-Score-Estimation-Tutorial

[James Hart]

Hi all,

The LD score files linked here by Dr. Grotzinger will work with the ldsc() function after some minor reformatting. The required reformatting is as follows: 1st, Seperate the *.rsid.l2.ldscore.gz files containing LD scores for all chromosomes into files for each chromosome, filenames should look like: 1.l2.ldscore.gz. for CHR 1. 2nd Generate a *.l2.M_5_50 file for each file, this is typically used for partitioned LD scores (which exist already for those UKB ld scores), in this case I just wrote the number of SNPs in each chr ld score file to file and saved like: 1.l2.M_5_50 for CHR 1.  Set ld and wld for the ldsc() function to the foulder containing these files. This worked for me using the African (AFR) ancestry.