ldsc39 ./munge_sumstats.py has issue with *.gz file

[YuandanZhang]

The new version ./munge_sumstats.py has issue with *.gz file eg GCST90475762.h.tsv.gz,
it's works with the decompressed file: eg GCST90475762.h.tsv

LD Score Regression (LDSC)
Version 3.0.1
(C) 2014-2019 Brendan Bulik-Sullivan and Hilary Finucane
Broad Institute of MIT and Harvard / MIT Department of Mathematics
GNU General Public License v3
Call:
./munge_sumstats.py
--sumstats GCST90475762.h.tsv.gz
--out GCST90475762.h.tsv.gz
--merge-alleles eur_w_ld_chr/w_hm3.snplist
--chunksize 50000
--snp rsid
--N-col n
--a1 effect_allele
--a2 other_allele
--p p_value
--frq effect_allele_frequency
--signed-sumstats odds_ratio,1
--ignore variant_id

ERROR converting summary statistics:

Traceback (most recent call last):
File "bin/ldsc3/munge_sumstats.py", line 534, in munge_sumstats
file_cnames = read_header(args.sumstats) # note keys not cleaned
File "bin/ldsc3/munge_sumstats.py", line 125, in read_header
return [x.rstrip('\n') for x in openfunc(fh).readline().split()]
File "bin/ldsc3/munge_sumstats.py", line 125, in
return [x.rstrip('\n') for x in openfunc(fh).readline().split()]
TypeError: a bytes-like object is required, not 'str'

Conversion finished at Sun May 24 22:14:27 2026
Total time elapsed: 0.03s

LD Score Regression (LDSC)
Version 3.0.1
(C) 2014-2019 Brendan Bulik-Sullivan and Hilary Finucane
Broad Institute of MIT and Harvard / MIT Department of Mathematics
GNU General Public License v3
Call:
./munge_sumstats.py
--sumstats GCST90475762.h.tsv
--out GCST90475762.h.tsv
--merge-alleles eur_w_ld_chr/w_hm3.snplist
--chunksize 50000
--snp rsid
--N-col n
--a1 effect_allele
--a2 other_allele
--p p_value
--frq effect_allele_frequency
--signed-sumstats odds_ratio,1
--ignore variant_id

{'effect_allele': 'A1', 'other_allele': 'A2', 'odds_ratio': 'SIGNED_SUMSTAT', 'effect_allele_frequency': 'FRQ', 'p_value': 'P', 'rsid': 'SNP', 'n': 'N'}
Interpreting column names as follows:
effect_allele: Allele 1, interpreted as ref allele for signed sumstat.
other_allele: Allele 2, interpreted as non-ref allele for signed sumstat.
odds_ratio: Directional summary statistic as specified by --signed-sumstats.
effect_allele_frequency: Allele frequency
p_value: p-Value
rsid: Variant ID (e.g., rs number)
n: Sample size

Reading list of SNPs for allele merge from eur_w_ld_chr/w_hm3.snplist
Read 1217311 SNPs for allele merge.
Reading sumstats from GCST90475762.h.tsv into memory 50000 SNPs at a time.
.......................................................................................................................................................................................................................................................................................................................................................................................................... done
Read 19668723 SNPs from --sumstats file.
Removed 18462319 SNPs not in --merge-alleles.
Removed 0 SNPs with missing values.
Removed 0 SNPs with INFO <= 0.9.
Removed 24038 SNPs with MAF <= 0.01.
Removed 0 SNPs with out-of-bounds p-values.
Removed 289 variants that were not SNPs or were strand-ambiguous.
1182077 SNPs remain.
Removed 499 SNPs with duplicated rs numbers (1181578 SNPs remain).
Removed 0 SNPs with N < 275902.6666666667 (1181578 SNPs remain).
Median value of SIGNED_SUMSTATS was 1.000100010001, which seems sensible.
Removed 164 SNPs whose alleles did not match --merge-alleles (1181414 SNPs remain).
Writing summary statistics for 1217311 SNPs (1181414 with nonmissing beta) to GCST90475762.h.tsv.sumstats.gz.

Metadata:
Mean chi^2 = 2.6
Lambda GC = 2.21
Max chi^2 = 1328.987
5202 Genome-wide significant SNPs (some may have been removed by filtering).

Conversion finished at Mon May 25 11:45:11 2026
Total time elapsed: 1.0m:50.05s

[quattro]

Replace read_header(fh) with

def read_header(fh):
      openfunc, _ = get_compression(fh)

      with openfunc(fh) as fileobj:
          line = fileobj.readline()

      if isinstance(line, bytes):
          line = line.decode("utf-8")

      return line.split()

[xiaoyao00]

Thanks for checking. The issue has been fixed by reading gzipped input headers in text mode.