bulik / ldsc

LD Score Regression (LDSC)
GNU General Public License v3.0
644 stars 343 forks source link

Which --signed-sumstats? And NA genetic correlation #320

Open jielab opened 3 years ago

jielab commented 3 years ago

Hi, guys:

I am still confused with 2 basic things when running LDSC.

  1. If my GWAS have BETA, SE, Z, OR, how should I specify --signed-sumstats? Should it be "BETA,0" or "Z,0" or "OR,1"? Will this affect my output *.sumstats.gz file and downstream LDSC analysis? I know some software such as GCTA requires "Z" but NOT "BETA".

  2. For the example below, why eur_w_ld_chr is specified twice? If I change the second one to "--w-ld-chr weights_hm3_no_hla/", I get IndexError: list index out of range. What is happening here? 11

  3. For the analysis below, why the heritability of phentoype 1 is negative, and the genetic correlation is NA NA NA.

image

Your clarification is appreciated!

Best regards, Jie

Soulnature commented 2 years ago

I have got the same question, have you solved the question yet? if done, could you give me some advice, thank you!

jielab commented 11 months ago

The following code works fine now:

conda activate ldsc dir=/mnt/d/data/gwas/penguin ldscdir=/mnt/d/data/ldsc dat=A dats="B C D E F G" datf=$dir/$dat.gz head_row=zcat $datf | head -1 | sed 's/\t/ /g'; munge_sumstats.py --chunksize 10000 --sumstats $dat.tmp.gz \ --merge-alleles $ldscdir/w_hm3.snplist --out $dat \ --snp SNP --a1 EA --a2 NEA --ignore SNPID,OR \ --signed-sumstats BETA,0 --p P --N 100000

echo $dat $dats | sed -e 's/ /.sumstats.gz,/g' -e 's/$/.sumstats.gz/' | \ xargs -n1 -I % ldsc.py --rg % --out $dat.rg \ --ref-ld-chr $ldscdir --w-ld-chr $ldscdir

Jie