$ operator is invalid for atomic vectors when running PUMA-ensemble.subsampling.R

[kvaddad1]

Hi Team,

Thank you so much for this great timely tool.

I got this error when running PUMA-ensemble.subsampling.R using the LD file downloaded from the official PUMAS repository.

Command used:
wget ftp://ftp.biostat.wisc.edu/pub/lu_group/Projects/PUMAS/LD/ld_1kg.RData

Rscript PUMA-ensemble.subsampling.R \
  --k 4 \
  --partitions 0.6,0.2,0.1,0.1 \
  --trait_name GCST90474619 \
  --ensemble EN \
  --gwas_path [my_gwas_dir] \
  --ld_path ld_1kg.RData \
  --output_path [my_output_dir]/PUMAS_height_output \
  --parallel \
  --threads 32

Error:
Error in match[[i]]$ld_blk : $ operator is invalid for atomic vectors
Calls: main -> match_gwas_LD
In addition: Warning message:
all scheduled cores encountered errors in user code
Execution halted

The GWAS file (GCST90474619.txt.gz) follows the expected QC format.

CHR    BP       SNP         A1  A2   MAF       BETA        SE          P               N
1      843942   rs4040617   G   A    0.158374  -0.0732193  0.03526765  0.03788436      2854
1      911428   rs4475691   T   C    0.216889   0.0035802  0.03028579  0.90589821      2854
.
.
.

It seems match_gwas_LD() is trying to access $ld_blk, but the LD object may be an atomic vector instead.
Could this be due to a format change or issue in the ld_1kg.RData file?
Kindly guide me on this.

[svdorn]

Hello, sorry about this issue!

Please use the 1000G_EUR.RData from this box folder and it should work, I just ran the subsampling step myself with this data to make sure. We reformatted the LD RData object which PUMAS accepts, and I need to remove the old download links from the README.

I updated the README now, so you can follow instructions from there as well!

[kvaddad1]

Hi @svdorn,

Thank you so much for your response and for sharing the box folder (https://uwmadison.app.box.com/s/c77uea6beiagxnkpa8rc2nagculgjcxa). It seems the institutional Box domain restricts direct downloads, and the link returns a 404 when accessed via command-line tools. Could you please provide an alternate mirror (Google Drive, Zenodo, or Globus) for the dataset?

Thanks again!