Genopred PGS rule issue

[shrivastavabhishek]

Hi Team,

I was trying generate the pgs scrore based on the external published GWAS sumstats data. Below is the format of input which I am using to run the genopred pipeline: This gwas sumstats has only 2955 number of SNPs.

CHR BP SNP A2 A1 N P BETA SE
6 32444703 chr6:32444703:G:A G A 19048 5.57E-14 -0.235722334 0.028910379
6 32444762 chr6:32444762:A:G A G 19048 5.57E-14 -0.235722334 0.028910379
6 32444794 chr6:32444794:A:T A T 19048 2.28E-12 -0.274436846 0.036744737

and is failing.

I have also check the overlap rsID with hammap_snplist and only 36 SNPs are overlapping.

I am not sure why it is breaking. Below are the error log for your review:

1. reading list of SNPs for allele merge from /tmp/RtmpW4Ovwt/hm3_no_mhc.snplist Read 1215912 SNPs for allele merge. Reading sumstats from /tmp/RtmpW4Ovwt/sumstats.gz into memory 5000000 SNPs at a time. . done Read 36 SNPs from --sumstats file. Removed 23 SNPs not in --merge-alleles. Removed 0 SNPs with missing values. Removed 0 SNPs with INFO <= 0.9. Removed 0 SNPs with MAF <= 0.01. Removed 0 SNPs with out-of-bounds p-values. Removed 0 variants that were not SNPs or were strand-ambiguous. 13 SNPs remain. Removed 0 SNPs with duplicated rs numbers (13 SNPs remain). Removed 0 SNPs with N < 12698.6666667 (13 SNPs remain). ERROR converting summary statistics: Traceback (most recent call last): File "/home/ext_shrivastav_abhishekkumar_may/CTD_PRS/resources/software/ldsc/munge_sumstats.py", line 701, in munge_sumstats check_median(dat.SIGNED_SUMSTAT, signed_sumstat_null, 0.1, sign_cname)

2. Error in rule ldsc_i: jobid: 8 input: /home/ext_shrivastav_abhishekkumar_may/GCA_PRS/GenoPredOut/reference/gwas_sumstat/GiantCellArthritis/GiantCellArthritis-cleaned.gz, /home/ext_shrivastav_abhishekkumar_may/CTD_PRS/resources/data/ld_scores, /home/ext_shrivastav_abhishekkumar_may/CTD_PRS/resources/software/ldsc/ldsc.py, /home/ext_shrivastav_abhishekkumar_may/CTD_PRS/resources/data/hm3_snplist/w_hm3.snplist output: /home/ext_shrivastav_abhishekkumar_may/GCA_PRS/GenoPredOut/reference/ldsc/GiantCellArthritis/GiantCellArthritis.hsq_obs log: /home/ext_shrivastav_abhishekkumar_may/GCA_PRS/GenoPredOut/reference/logs/ldsc-GiantCellArthritis.log (check log file(s) for error details) conda-env: /tools/GenoPred/pipeline/.snakemake/conda/a5a8a0f260443effe6ae6b9313150c49_ shell: Rscript ../Scripts/ldsc/ldsc.R --ref_plink_chr /home/ext_shrivastav_abhishekkumar_may/CTD_PRS/resources/data/ref/ref.chr --sumstats /home/ext_shrivastav_abhishekkumar_may/GCA_PRS/GenoPredOut/reference/gwas_sumstat/GiantCellArthritis/GiantCellArthritis-cleaned.gz --munge_sumstats /home/ext_shrivastav_abhishekkumar_may/CTD_PRS/resources/software/ldsc/munge_sumstats.py --ldsc /home/ext_shrivastav_abhishekkumar_may/CTD_PRS/resources/software/ldsc/ldsc.py --ld_scores /home/ext_shrivastav_abhishekkumar_may/CTD_PRS/resources/data/ld_scores/UKBB.EUR.rsid --hm3_snplist /home/ext_shrivastav_abhishekkumar_may/CTD_PRS/resources/data/hm3_snplist/w_hm3.snplist --hm3_no_mhc T --sample_prev 0.5 --pop_prev 0.1 --output /home/ext_shrivastav_abhishekkumar_may/GCA_PRS/GenoPredOut/reference/ldsc/GiantCellArthritis/GiantCellArthritis --test NA > /home/ext_shrivastav_abhishekkumar_may/GCA_PRS/GenoPredOut/reference/logs/ldsc-GiantCellArthritis.log 2>&1 (one of the commands exited with non-zero exit code; note that snakemake uses bash strict mode!) [Wed Feb 18 22:29:52 2026] Finished job 1. 2 of 6 steps (33%) done [Wed Feb 18 22:55:04 2026] Finished job 14. 3 of 6 steps (50%) done Shutting down, this might take some time. Exiting because a job execution failed. Look above for error message Complete log: .snakemake/log/2026-02-18T222940.351253.snakemake.log

3. snakemake log:
   output: /home/ext_shrivastav_abhishekkumar_may/GCA_PRS/GenoPredOut/reference/ldsc/GiantCellArthritis/GiantCellArthritis.hsq_obs
   log: /home/ext_shrivastav_abhishekkumar_may/GCA_PRS/GenoPredOut/reference/logs/ldsc-GiantCellArthritis.log
   jobid: 8
   benchmark: /home/ext_shrivastav_abhishekkumar_may/GCA_PRS/GenoPredOut/reference/benchmarks/ldsc-GiantCellArthritis.txt
   reason: Missing output files: /home/ext_shrivastav_abhishekkumar_may/GCA_PRS/GenoPredOut/reference/ldsc/GiantCellArthritis/GiantCellArthritis.hsq_obs; Input files updated by another job: /home/ext_shrivastav_abhishekkumar_may/GCA_PRS/GenoPredOut/reference/gwas_sumstat/GiantCellArthritis/GiantCellArthritis-cleaned.gz
   wildcards: gwas=GiantCellArthritis
   resources: tmpdir=/tmp

Activating conda environment: .snakemake/conda/a5a8a0f260443effe6ae6b9313150c49_
[Wed Feb 18 22:29:48 2026]
Error in rule ldsc_i:
jobid: 8
input: /home/ext_shrivastav_abhishekkumar_may/GCA_PRS/GenoPredOut/reference/gwas_sumstat/GiantCellArthritis/GiantCellArthritis-cleaned.gz, /home/ext_shrivastav_abhishekkumar_may/CTD_PRS/resources/data/ld_scores, /home/ext_shrivastav_abhishekkumar_may/CTD_PRS/resources/software/ldsc/ldsc.py, /home/ext_shrivastav_abhishekkumar_may/CTD_PRS/resources/data/hm3_snplist/w_hm3.snplist
output: /home/ext_shrivastav_abhishekkumar_may/GCA_PRS/GenoPredOut/reference/ldsc/GiantCellArthritis/GiantCellArthritis.hsq_obs
log: /home/ext_shrivastav_abhishekkumar_may/GCA_PRS/GenoPredOut/reference/logs/ldsc-GiantCellArthritis.log (check log file(s) for error details)
conda-env: /tools/GenoPred/pipeline/.snakemake/conda/a5a8a0f260443effe6ae6b9313150c49_
shell:
Rscript ../Scripts/ldsc/ldsc.R --ref_plink_chr /home/ext_shrivastav_abhishekkumar_may/CTD_PRS/resources/data/ref/ref.chr --sumstats /home/ext_shrivastav_abhishekkumar_may/GCA_PRS/GenoPredOut/reference/gwas_sumstat/GiantCellArthritis/GiantCellArthritis-cleaned.gz --munge_sumstats /home/ext_shrivastav_abhishekkumar_may/CTD_PRS/resources/software/ldsc/munge_sumstats.py --ldsc /home/ext_shrivastav_abhishekkumar_may/CTD_PRS/resources/software/ldsc/ldsc.py --ld_scores /home/ext_shrivastav_abhishekkumar_may/CTD_PRS/resources/data/ld_scores/UKBB.EUR.rsid --hm3_snplist /home/ext_shrivastav_abhishekkumar_may/CTD_PRS/resources/data/hm3_snplist/w_hm3.snplist --hm3_no_mhc T --sample_prev 0.5 --pop_prev 0.1 --output /home/ext_shrivastav_abhishekkumar_may/GCA_PRS/GenoPredOut/reference/ldsc/GiantCellArthritis/GiantCellArthritis --test NA > /home/ext_shrivastav_abhishekkumar_may/GCA_PRS/GenoPredOut/reference/logs/ldsc-GiantCellArthritis.log 2>&1
(one of the commands exited with non-zero exit code; note that snakemake uses bash strict mode!)
Any help is highly appreciable
3.

[opain]

Hello,

Thank you for the detailed report.

LDSC is appears to be failing due to the low number of non-MHC variants in your GWAS. Have you subset the GWAS to a specific locus?

You could try adding this parameter to the config file, so it uses a different method to estimate SNP-h2:

h2_method: 'sbayesr'

Only DBSLMM and LDpred require SNP-based heritability estimates in advance, so you could avoid using these methods. pt_clump should have no issues. However, other issues may arise. The pipeline is primarily set up for genome-wide GWAS sumstats.

Best wishes,

Ollie

[shrivastavabhishek]

Hi Ollie,

Thanks for your guidance and h2_method: 'sbayesr' worked and it completed without any error.

Yes, I am using the subset of sumstats and that is limited 2955 number of SNPs. Do we also need to be careful about genomebuild before running the Genopred if we are including dbSLMM in the method and can we replace hm3_snplist with our own snp list to calculate heritability as these number are very limited.

I am also curious, based on overlapped of only 36 SNPs at Sbayes steps how it is calculating the PRS for all the other SNPs located on other chromosome. I am getting around 809 SNPs with Polygenic risk score across other chromosme.
Any insight will be highly appreciated.

However, I considerd only sbayes and pt_clump method in config file.

Thank you,
Abhishek

[opain]

Hi Abhishek,

Thanks for the information.

Genome build shouldn't be an issue - If using CHR and BP to match variants to the reference, it first detects which build the target/gwas data is in. If it is using RSIDs, then build does not matter.

You cannot expand the list of variants used by LDSC beyond hapmap3 variants, as the LD scores are restricted to hapmap3 variants. You can provide your own expanded reference data to be used by GenoPred for other pats of the pipeline though. Here are some instructions: https://opain.github.io/GenoPred/pipeline_readme.html#Using_your_own_reference and https://opain.github.io/GenoPred/pipeline_readme.html#Specifying_alternative_reference_data_for_PGS_methods

There might be fewer variants overlapping between the GWAS and the reference used by SBayesR, as the SBayesR reference was created by the software developers and may not overlap completely with the reference data GenoPred uses for other steps. Unless you are using SBayesRC, the score files output by PGS methods should only contain variants that were present in your GWAS summary statistics.

Suggestion: I assume you are trying to create a partitioned polygenic score, restricted to a certain set of variants. Since GenoPred and most PGS methods are designed for genome-wide GWAS sumstats, I would suggest using the full sumstats as the input for GenoPred, and then subsetting the PGS score files to the variants you are interested in later when calculating the PGS in the target sample. I hope to release better functionality for partitioned PGS later this year.

Best wishes,

Ollie

[shrivastavabhishek]

Hi Ollie,

Thanks for the reply.
I went through some of the log files and results where I was able to find some other SNPs which was not part of my input gwas sum stats files and only 36 SNPs was overlapping with input gwas sum ststas with score files.

I also went through "GenoPredTest/reference/gwas_sumstat" directory where I found the log where it is creating TargetGCA-cleaned file. here it is able to map only 36 SNPs with ref file(resources/data/ref/ref.chr).

log for gwas sumstats:
After removal of SNPs with reference MAF < 0.01, 36 variants remain.
Reported MAF column is not present, so discordance with reference cannot be determined.
After removal of SNPs with out-of-bound P values, 36 variants remain.
After removal of SNPs with duplicate IDs, 36 variants remain.
N column is not present or invariant.
Genomic control was not detected.
After removal of SNPs with SE == 0, 36 variants remain.

and then this clean gwas_sumstats is being used by sbayesRC where again the score is only for same 36 SNPs.

pgs_score_files/sbayesrc/TargetGCA$ ref-TargetGCA.log details:
Parameter Value
ref_plink_chr /home/ext_shrivastav_abhishekkumar_may/GCA_PRS/resources/data/ref/ref.chr
ref_pcs /home/ext_shrivastav_abhishekkumar_may/GCA_PRS/GenoPredTest/reference/pc_score_files/TRANS/ref-TRANS-pcs.profiles
pop_data /home/ext_shrivastav_abhishekkumar_may/GCA_PRS/resources/data/ref/ref.pop.txt
plink2 plink2
output /home/ext_shrivastav_abhishekkumar_may/GCA_PRS/GenoPredTest/reference/pgs_score_files/sbayesrc/TargetGCA/ref-TargetGCA
n_cores 10
sumstats /home/ext_shrivastav_abhishekkumar_may/GCA_PRS/GenoPredTest/reference/gwas_sumstat/TargetGCA/TargetGCA-cleaned.gz
gctb /home/ext_shrivastav_abhishekkumar_may/GCA_PRS/resources/software/gctb_2.5.2/gctb_2.5.2_Linux/gctb
test NA
seed 1
sbayesrc_ldref /home/ext_shrivastav_abhishekkumar_may/GCA_PRS/resources/data/sbayesrc_ref/EUR
sbayesrc_annot /home/ext_shrivastav_abhishekkumar_may/GCA_PRS/resources/data/sbayesrc_annot/annot_baseline2.2.txt
help FALSE
output_dir /home/ext_shrivastav_abhishekkumar_may/GCA_PRS/GenoPredTest/reference/pgs_score_files/sbayesrc/TargetGCA/

Analysis started at 2026-02-23 17:28:18.70715
Reading in GWAS.
sumstats contains 36 variants.
REF.FREQ being used as FREQ.
sumstats contains 36 variants with complete data.
23 variants remain after SBayesRC tidy step.
Running SBayesRC imputation step...
Running SBayesRC main analysis...
Analysis finished at 2026-02-23 17:53:39.125431
Analysis duration was 25.34 mins

It would be great if you can recommend how we can get pgs score for only our inputs gwas sumstas file not created by genopred during the workflow.

Your help is highly appreciated.

Thanking you,
Abhishek

[opain]

Hi Abhishek,

SBayesRC is different to other PGS methods implemented in GenoPred, as it tries to impute missing variants. This is because SBayesRC uses eigen decomposed LD data, thereby requiring all variants in its reference to be present when modelling SNP effects.

Any other PGS method in GenoPred will be restricted to variants in the GWAS.

I hope this clarifies things.

Best wishes,

Ollie