bulik / ldsc

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

Partitioned heritability: Error parsing reference panel LD Score. #223

Open geneticsmcgill opened 4 years ago

geneticsmcgill commented 4 years ago

Hi there, I keep getting an error for partitioned heritability. I've tried multiple different combinations of files (including the suggested ones in the tutorial) and with or without merging with hm3 dataset prior. Any help would be appreciated.

Error below:

Call: ./ldsc.py \ --h2 munged.rsid.filtered.merged_METAANALYSIS.ETGWAS.inclUKBB_2020-06-16.sumstats.gz \ --ref-ld-chr ./1000G_Phase3_baselineLD_v2.2_ldscores/baselineLD. \ --out partictionedh2.munged.hm3.rsid.filtered.merged_METAANALYSIS.ETGWAS.inclUKBB_2020-06-16_ldhub.txt.sumstats \ --overlap-annot \ --frqfile-chr ./plink_files/1000G.EUR.hg38. \ --w-ld-chr weights/weights.hm3_noMHC.

Beginning analysis at Mon Jun 22 19:54:15 2020 Reading summary statistics from munged.rsid.filtered.merged_METAANALYSIS.ETGWAS.inclUKBB_2020-06-16.sumstats.gz ... Read summary statistics for 5850543 SNPs. Reading reference panel LD Score from ./1000G_Phase3_baselineLD_v2.2_ldscores/baselineLD.[1-22] ... Error parsing reference panel LD Score. Traceback (most recent call last): File "./ldsc.py", line 644, in sumstats.estimate_h2(args, log) File "/Users/cl/ldsc/ldscore/sumstats.py", line 329, in estimate_h2 args, log, args.h2) File "/Users/cl/ldsc/ldscore/sumstats.py", line 246, in _read_ld_sumstats ref_ld = _read_ref_ld(args, log) File "/Users/cl/ldsc/ldscore/sumstats.py", line 85, in _read_ref_ld 'reference panel LD Score', ps.ldscore_fromlist) File "/Users/cl/ldsc/ldscore/sumstats.py", line 158, in _read_chr_split_files raise e ValueError: No objects to concatenate

Analysis finished at Mon Jun 22 19:54:25 2020 Total time elapsed: 10.02s Traceback (most recent call last): File "./ldsc.py", line 644, in sumstats.estimate_h2(args, log) File "/Users/cl/ldsc/ldscore/sumstats.py", line 329, in estimate_h2 args, log, args.h2) File "/Users/cl/ldsc/ldscore/sumstats.py", line 246, in _read_ld_sumstats ref_ld = _read_ref_ld(args, log) File "/Users/cl/ldsc/ldscore/sumstats.py", line 85, in _read_ref_ld 'reference panel LD Score', ps.ldscore_fromlist) File "/Users/cl/ldsc/ldscore/sumstats.py", line 158, in _read_chr_split_files raise e ValueError: No objects to concatenate

quattro commented 4 years ago

I think the glob call in sumstats.py::_splitp is the culprit. It should append ".*"

futurologist commented 4 years ago

Hi, I also encountered the same problem. For me, the current version of ldsc produces exactly the same error when running the tutorial examples. Basically, it works for heritability and genetic correlation, but fails the tutorial suggested tests for partitioned heritability:

Beginning analysis at Fri Jun 26 15:32:41 2020 The frequency file is unnecessary and is being ignored. Reading summary statistics from /home/LDSC_work/outs_part_herit_sumstats/BMI.sumstats.gz ... Read summary statistics for 1040803 SNPs. Reading reference panel LD Score from /home/LDSC_work/dwnlds_partition_herit/baseline/baseline.[1-22] ... Error parsing reference panel LD Score. Traceback (most recent call last): File "/home/ldsc/ldsc.py", line 644, in sumstats.estimate_h2(args, log) File "/home/ldsc/ldscore/sumstats.py", line 329, in estimate_h2 args, log, args.h2) File "/home/ldsc/ldscore/sumstats.py", line 246, in _read_ld_sumstats ref_ld = _read_ref_ld(args, log) File "/home/ldsc/ldscore/sumstats.py", line 85, in _read_ref_ld 'reference panel LD Score', ps.ldscore_fromlist) File "/home/ldsc/ldscore/sumstats.py", line 158, in _read_chr_split_files raise e ValueError: No objects to concatenate

After following the suggestions from previous discussions, I made LDSC work by simply going back to an older version, before the current error was introduced.

If one downloads LDSC for the first time:

$ git clone https://github.com/bulik/ldsc.git $ cd ldsc $ git checkout b02f2a6c6d4c2724a7d88eca8a85c952762e8f0f

annashcherbina commented 4 years ago

Same issue here on tutorial datasets and my own data -- confirmed that reverting to b02f2a6 fixes the problem.

rodrigoduarte88 commented 4 years ago

git checkout b02f2a6 did the work for me! Thanks!