PRSet: Pathway-based polygenic risk score analyses and software

Abstract

Polygenic risk scores (PRSs) have been among the leading advances in biomedicine in recent years. As a proxy of genetic liability, PRSs are utilised across multiple fields and applications. While numerous statistical and machine learning methods have been developed to optimise their predictive accuracy, these typically distil genetic liability to a single number based on aggregation of an individual’s genome-wide risk alleles. This results in a key loss of information about an individual’s genetic profile, which could be critical given the functional sub-structure of the genome and the heterogeneity of complex disease. In this manuscript, we introduce a ‘pathway polygenic’ paradigm of disease risk, in which multiple genetic liabilities underlie complex diseases, rather than a single genome-wide liability. We describe a method and accompanying software, PRSet, for computing and analysing pathway-based PRSs, in which polygenic scores are calculated across genomic pathways for each individual. We evaluate the potential of pathway PRSs in two distinct ways, creating two major sections: (1) In the first section, we benchmark PRSet as a pathway enrichment tool, evaluating its capacity to capture GWAS signal in pathways. We find that for target sample sizes of >10,000 individuals, pathway PRSs have similar power for evaluating pathway enrichment as leading methods MAGMA and LD score regression, with the distinct advantage of providing individual-level estimates of genetic liability for each pathway–opening up a range of pathway-based PRS applications, (2) In the second section, we evaluate the performance of pathway PRSs for disease stratification. We show that using a supervised disease stratification approach, pathway PRSs (computed by PRSet) outperform two standard genome-wide PRSs (computed by C+T and lassosum) for classifying disease subtypes in 20 of 21 scenarios tested. As the definition and functional annotation of pathways becomes increasingly refined, we expect pathway PRSs to offer key insights into the heterogeneity of complex disease and treatment response, to generate biologically tractable therapeutic targets from polygenic signal, and, ultimately, to provide a powerful path to precision medicine.

Non-additive effects of schizophrenia risk genes reflect convergent downstream function

“…Specifically, we tested (1) whether non-additive genes are enriched in schizophrenia GWAS signal, and (2) to what extent non-additive genes are part of the total genetic risk load to schizophrenia. Pathway PRS for non-additive genes from each functional category were calculated using PRSet 42 (see Methods ), and results were compared with standard, genome-wide PRS (PRSice) 43 and with pathway PRS from curated lists of 1233 synaptic genes (Synapse Gene Ontology 44 ) and 1639 transcription factors (The Human Transcription Factor database 45 ).…”

“…Our data argues for a re-conceptualization of genetic contribution to disease risk; PRSs that incorporate biological pathways and gene co-expression networks of the risk variants being summed 42,71 , together with their downstream impacts 47 , may improve patient stratification 42,72 or better predict drug response 73 . Moreover, resolving convergent effects shared between eGenes will identify nodes whereby the impact of aggregate genetic risk can be most effectively decreased and prioritize potential therapeutic targets.…”

Dopamine and schizophrenia from bench to bedside: Discovery of a striatal co-expression risk gene set that predicts in vivo measures of striatal function

“…To stratify SCZ genetic risk for genes within a specific component we first mapped European 1000G SNPs at 100kbp up- and down-stream of each component-specific gene using MAGMA tool (v1.09b), we then matched component-specific SNPs with PGC3 leave-LIBD-out summary statistics and finally computed the scores for the KCL, NIMH, LIBD- and UNIBA-fMRI cohort separately using PRSset 96 and again the European 1000G as LD reference panel. As negative control, we computed complementary scores including all PGC3 leave-LIBD-out SNPs not mapping to any of the component-specific genes.…”

Pathway-Specific Polygenic Scores Improve Cross-Ancestry Prediction of Psychosis and Clinical Outcomes

“…In addition to the overall PGS, we calculated pPGS for the nine gene ontology pathways which represented independent enrichment signals in the most recent GWAS of SCZ (Trubetskoy et al, 2022): axon, ion channel complex, nervous system development, neuronal cell body, somatodendritic compartment, synapse, regulation of cation channel activity, regulation of neuron differentiation, voltage-gated calcium channel activity. In order to ensure complete coverage of pathway genes, a clumping and thresholding approach implemented in the PRSet software package (Choi et al, 2023) was used to calculate pPGS for the nine biological pathways of interest. For pPGS calculation, the 1000 genomes European reference sample was used for LD calculation and a clumping R 2 threshold of 0.1 was applied, which were found to optimize prediction of case-control status in the target sample using a genome-wide SCZ PGS calculated by clumping and thresholding.…”