search

bulik / ldsc

LD Score Regression (LDSC)
GNU General Public License v3.0
634 stars 342 forks source link

Suitable w_hm3.snplist for AFR? #407

Open beckse opened 11 months ago

beckse commented 11 months ago

Hi! I've come across a LD reference panel for AFR. But I need an appropriate SNP list for --merge-alleles. Does anyone know how to generate something suitable? Thank you!

samreenzafer commented 11 months ago

Can I ask where you find it? I need a ref panel for AFR population also for using with MTAG (which relies on LDSC).

laleoarrow commented 4 months ago

Also need the w_hm3.snplist file for non-EUR population. Any resources on this matter is highly appreciated!

aksarkar commented 4 months ago

@leoarrow1 Use the MAP files from HapMap 3 for your population of interest to get the variants to include https://www.broadinstitute.org/files/shared/mpg/hapmap3/hapmap3_r1_b36_fwd.qc.poly.tar.bz2

If you want superpopulations (for example, AFR), take the union of the constituent population MAP files.

The population codes are given at https://www.broadinstitute.org/medical-and-population-genetics/hapmap-3

You will need to lift over positions/IDs from b36 to match your reference LD matrices.

You will then need to compute the weights yourself as (non-partitioned) LD scores from your matched reference panel.

yuj23 commented 4 months ago

@leoarrow1 Use the MAP files from HapMap 3 for your population of interest to get the variants to include https://www.broadinstitute.org/files/shared/mpg/hapmap3/hapmap3_r1_b36_fwd.qc.poly.tar.bz2

If you want superpopulations (for example, AFR), take the union of the constituent population MAP files.

The population codes are given at https://www.broadinstitute.org/medical-and-population-genetics/hapmap-3

You will need to lift over positions/IDs from b36 to match your reference LD matrices.

You will then need to compute the weights yourself as (non-partitioned) LD scores from your matched reference panel.

Do you know how to get those snps in w_hm3.snplist? The ldscore in w_ld_eur/ does not fully match the w_hm3.snplist list. Is w_hm3.snplist only for EUR population?

laleoarrow commented 4 months ago

@yuj23 According to this #425 issue, it seems that w_hm3.snplist is not EUR specific and can be applied to other ancestry.

aksarkar commented 4 months ago

@yuj23 From Bulik-Sullivan et al. 2015: "In order to restrict to a set of common, well-imputed variants, we retained only those SNPs in the HapMap 3 reference panel for the LD Score regression."

These would likely be https://www.broadinstitute.org/files/shared/mpg/hapmap3/hapmap3_r1_b36_fwd_consensus.qc.poly.recode.map.bz2 which are consensus calls over all HapMap subpopulations.