Estimating LD scores

This page describes the steps to estimate LD scores from LD reference panels using S-LDXR.

Running S-LDXR to estimate LD scores

S-LDXR estimates LD scores with the following command.

for chrom in $(seq 22)
do
    python <software directory>/s-ldxr.py \
        --score allelic \
        --ld-wind-cm 1.0 \
        --print-snps <a list of SNPs to print> \
        --bfile <EAS reference panel directory>/1000G.EAS.${chrom} \
                <EUR reference panel directory>/1000G.EUR.${chrom} \
        --annot <annotation directory>/<annotation file>.${chrom}.annot.gz \
        --out <output directory>/<ld score file prefix>.${chrom}
done
Note: The above for loop can be parallelized by chromosome.

Here are the meanings of the flags:

  • --score specifies the type of LD scores to be estimated. Here, one should almost always put "allelic" as the score type for estimating per-allele effect correlation. Although "standardized" score type is also supported, we do not recommend to only use it in computing regression weights, as trans-ethnic genetic correlation of standardized causal effect sizes is not very interpretable.

  • --ld-wind-cm specifies the maximum window size in centimorgan for estimating LD scores. The default and recommended value is 1.0.

  • --print-snps specifies the file that contains a list of SNPs in plain text without header, for which the LD scores are to be printed. We recommend to print all SNPs with minor allele frequency greater than 1% in both populations.

  • --bfile takes two argument -- reference panel for population 1, and reference panel for population 2. All reference panels should be in PLINK format, and have the same set of SNPs.

  • --annot specifies the annotation file.

  • --out specifies prefix of the output files.

Output of S-LDXR

After executing the command above, S-LDXR creates 4 files for each chromosome (i.e. 88 files for all 22 chromosomes):

  • <ld score file prefix>.<chrom>_pop1.gz - LD score files for the 1st population (correspond to EAS in the above command).

  • <ld score file prefix>.<chrom>_pop2.gz - LD score files for the 2nd population (correspond to EUR in the above command).

  • <ld score file prefix>.<chrom>_te.gz - Trans-ethnic LD score files.

  • <ld score file prefix>.<chrom>.log - contains helpful information for debugging, including number of SNPs, number of SNPs filtered, etc.

Note: Log files are very useful in pinpointing bugs of the software. Please include the log file in the email in any bug report.