PRS Scoring File Low Overlap

[Lord-of-Bugs]

Hi there,

Would appreciate someone help pointing out the issue here. I am trying to prepare my own sample individual data for calculating PRS. Here's how I prepared my data:

1. Map my genotyping results to the variants present on Illumina's GSA v3's manifest file (https://support.illumina.com/downloads/infinium-global-screening-array-v3-0-product-files.html), using the one that's build 38.
2. Extract the called markers based on chromosome and place them into different VCF files (so I made 22 of them).
3. Prepare the samplesheet accordingly.
4. Run PRS calculator (I'm trying to calculate scores for PGS000001 and PGS000004).

My nextflow pipeline is set up just fine. But every time the pipeline breaks down before neither score passes the match rate threshold:

  pgscatalog.match.lib.matchresult: 2025-04-25 10:28:47 WARNING  Score PGS000004_hmPOS_GRCh38 matching failed with match rate 0.09904153354632583
  pgscatalog.match.lib.matchresult: 2025-04-25 10:28:47 WARNING  Score PGS000001_hmPOS_GRCh38 matching failed with match rate 0.4545454545454546

I ran a manual inspection about my data. Basically, I grabbed the rs IDs present in the PGS000001 scoring file and checked for existence in the GSA v3 build 38 manifest, and it seems that > 90% of the markers in the scoring file are present in my data.

For example, here's my called markers matched to the manifest file and processed to be almost like a VCF (df):

#CHROM	POS	ID	REF	ALT	QUAL	FILTER	INFO	FORMAT	HG01529
1	102914837	1:103380393	T	C	.	PASS	.	GT	1|1
1	108897058	1:109439680	A	G	.	PASS	.	GT	0|0
1	109656166	1:110198788	T	C	.	PASS	.	GT	0|0
1	109658490	1:110201112	G	C	.	PASS	.	GT	0|0
1	109659045	1:110201667	T	C	.	PASS	.	GT	1|1

Scoring file dataframe (breast_cancer_df):

rsID	chr_name	effect_allele	other_allele	effect_weight	locus_name	OR
rs78540526	11	T	C	0.162204	CCND1	1.1761
rs75915166	11	A	C	0.023619	CCND1	1.0239
rs554219	11	G	C	0.116716	CCND1	1.1238
rs7726159	5	A	C	0.035271	TERT	1.0359
rs10069690	5	T	C	0.023912	TERT	1.0242

And

>>> print(breast_cancer_df[breast_cancer_df['rsID'].isin(df['ID'])].shape[0] / breast_cancer_df.shape[0])
0.922

So now I am completely unsure why the pipeline does not meet the overlap threshold. I ran the following commands:

nextflow run pgscatalog/pgsc_calc -profile docker,arm -resume -r v2.0.0 \
--input ./samplesheet.csv \
--target_build GRCh38 \
--pgs_id PGS000001, PGS000004 \
--genotypes_cache cache \
--run_ancestry ./pgsc_HGDP+1kGP_v1.tar.zst \
--outdir result

Would greatly appreciate someone taking a stab at this!!!

[smlmbrt]

The matching doesn't work on rsIDs, it works on chromosome, position, and then the alleles. Could you extract some matched rows from the df and the vcf and some that didn't? Might be able to help that way.