PGS Catalog Calculator (pgsc_calc)
Summary¶
Lambert, Wingfield, Gibson, et al. (2024) present PGS Catalog Calculator (pgsc_calc), an open-source Nextflow pipeline that reproducibly calculates PGS Catalog scores (or user-supplied custom scoring files) on individual-level genomic data at biobank scale, and applies genetic-ancestry-aware normalization so PGS distributions become comparable across ancestry groups — addressing the previous lack of any standalone, portable software for either of these two steps. Prior to this tool, applying PGS Catalog scoring files required manually chaining together separate genotype-formatting, variant-matching, and allelic-scoring tools that varied across platforms, and no dedicated ancestry-adjustment package existed.
Capabilities¶
- Variant standardization: automatically matches, lifts over (GRCh37/GRCh38), and aligns alleles between the target genotype dataset and PGS scoring files.
- Ancestry calibration: implements the most common PGS-ancestry-adjustment methods (Box 1 of the paper) using a reference panel (e.g., HGDP+1kGP) — including discrete "most similar population" normalization (Z_MostSimilarPop) and continuous mean/variance adjustment methods (Z_norm1, Z_norm2) that do not require discrete population labels.
- Nextflow backend: built on Nextflow/nf-core for portability across local machines, HPC clusters (SLURM), cloud platforms (AWS, Google Cloud), and airlocked trusted research environments, using standard reproducible-software distribution (Docker, Singularity, Conda) — designed explicitly to bring "code to the data" rather than requiring genetic data upload to a web server.
- Flexible score specification: users can request scores by PGS Catalog score/publication/trait identifier, or supply their own custom scoring files; results are aggregated into a tabular output file plus a summary report describing calculated scores, variant matching, and PGS distributions.
Demonstrated Ancestry Adjustment¶
Applying pgsc_calc to PGS000018 (metaGRS_CAD, 1,745,179 variants) across ~487,000 UK Biobank individuals with an HGDP+1kGP reference panel showed that raw weighted-sum PGS distributions differ substantially in mean and variance across ancestry superpopulations (African, American, Central/South Asian, East Asian, European) — differences that reflect population-level allele frequency and LD structure rather than true differences in disease risk. The pipeline's ancestry-adjustment methods (Z_MostSimilarPop, Z_norm1, Z_norm2) made these distributions substantially more comparable across ancestry groups, illustrating the practical necessity of ancestry-aware normalization for equitable PGS interpretation.
Availability¶
Code and documentation: github.com/PGScatalog/pgsc_calc (readthedocs).
See Also¶
- PGS Catalog — the open database of scoring files this calculator applies to individual-level data.
- OmicsPred — a related resource of genetic prediction models this calculator's scoring infrastructure is compatible with.