LD Scores for custom annotations not matching baseline model SNPs

[tmaj...@broadinstitute.org]

Hi-

I have generated some genomewide annotations that I would like to incorporate with the "baseline" model in partitioning heritability. I have followed the instructions for generating annotation files and LD scores using 1000G Phase 3 data and HapMap 3 snps found here: https://github.com/bulik/ldsc/wiki/LD-Score-Estimation-Tutorial

All of this seemed to work until I tried partitioning heritability. I am getting the following error: "ValueError: LD Scores for concatenation must have identical SNP columns."

Looking into the discordance between the two LD score files (baseline and my own), I see that the two files do have some differences. Neither file is a superset of the other but rather, each has some SNPs not included in the other.

For example: 

bt <- fread("1000G_EUR_Phase3_baseline/baseline.19.l2.ldscore.gz", data.table = F, stringsAsFactors = F)

rv <- fread("rarevar.1000G.EUR.QC.19.l2.ldscore", data.table = F, stringsAsFactors = F)

head(bt$SNP)

[1] "rs8100066"  "rs8105536"  "rs2312724"  "rs1020382"  "rs12459906" "rs11084928"

head(rv$SNP)

[1] "rs8100066"  "rs8102615"  "rs8105536"  "rs2312724"  "rs1020382"  "rs12459906"

sum(!(bt$SNP %in% rv$SNP))

[1] 2051

sum(!(rv$SNP %in% bt$SNP))

[1] 1533

I used the plink files and hapmap snps recommended in the documentation: 1000G_Phase3_plinkfiles.tgz, hapmap3_snps.tgz.

Any ideas as to why I'm not getting the same snps would be appreciated.

Thanks.

[leihou...@gmail.com]

I also found the same problem, and I used baseline annotations from folder 1000G_Phase3_baselineLD_v2.1_ldscores.

do you get anything back from them?

[tmaj...@broadinstitute.org]

No, have not heard back.

[Paul Hook]

Hello all,

I ran into this same issue with "1000G_Phase3_baseline_ldscores.tgz"

Within that baseline folder was a file called "print_snps.txt" which (it seems) was used to create the baseline LD scores. This seemed to be a different list of SNPs than those included in the HapMap SNPs referred to in the tutorial (https://github.com/bulik/ldsc/wiki/LD-Score-Estimation-Tutorial).

I got around this issue by reprocessing the baseline BED files with the HapMap SNPs.

Paul

[Koen]

Hi Paul,

I was wondering if you have any scripts available to reprocess the BED files?

Best,

Koen

Op maandag 18 februari 2019 16:16:29 UTC+1 schreef Paul Hook: