Traditional genetics often searches for a single broken gear—a high-impact monogenic mutation like BRCA1. However, complex traits such as depression, schizophrenia, and coronary artery disease do not follow this simple blueprint. Instead, they are polygenic, shaped by the collective influence of thousands of single-nucleotide polymorphisms (SNPs) scattered across the genome. In isolation, each variant has a microscopic, statistically insignificant effect. But when aggregated, their signals build a powerful cumulative picture of genetic liability.
As explained by the authors of Embracing polygenicity: a review of methods and tools for psychiatric genetics research, polygenic risk scoring collapses this high-dimensional genomic data into a single continuous value. This approach captures the cumulative impact of small-effect variants that traditional single-variant analyses miss. In fact, Natarajan et al. demonstrate in Deep-coverage whole genome sequences and blood lipids among 16,324 individuals that for conditions like severe hypercholesterolemia, individuals are roughly ten times more likely to be driven by a high cumulative polygenic burden than by a single high-impact monogenic mutation.
The Mechanics of Aggregation
The fundamental mechanism behind a polygenic score is an additive model. For a given individual, we count the number of risk alleles (0, 1, or 2) across thousands of genetic loci and multiply each by a "weight"—the effect size derived from an independent, large-scale Genome-Wide Association Study (GWAS).
This basic framework allows researchers to explore how broad genetic predispositions manifest in specific clinical outcomes. For example, Amare et al. use this strategy in Association of the Polygenic Scores for Personality Traits and Response to Selective Serotonin Reuptake Inhibitors in Patients with Major Depressive Disorder to show that cumulative genetic indicators for personality traits like conscientiousness can stratify patients by their likelihood of early antidepressant success.
However, simple summation faces a major biological hurdle: linkage disequilibrium (LD), the tendency for nearby genetic variants to be inherited together. If uncorrected, LD leads to double-counting correlated markers and distorting the score. To resolve this, advanced models employ Bayesian frameworks to account for LD structures and more accurately model the true distribution of effect sizes. As shown by Yap et al. in Analysis of common genetic variation and rare CNVs in the Australian Autism Biobank, Bayesian polygenic scoring significantly improves the precision of risk estimates compared to standard summation techniques. Similarly, Dennis et al. leverage these probabilistic models in Clinical laboratory test-wide association scan of polygenic scores identifies biomarkers of complex disease to clean up noisy summary statistics and map genetic liability directly to clinical biomarkers.
Disentangling Biology via Pathway and Expression Scores
While genome-wide scores provide a comprehensive view of overall liability, they treat the genome as a black box, blending thousands of distinct biological processes together. To extract clearer mechanistic insights, researchers construct pathway-specific polygenic scores. By restricting the aggregation of variants to only those located within or near genes belonging to a specific biological pathway, they can isolate the genetic burden on a single molecular system.
For instance, Barbu et al. demonstrate this in Association of Whole-Genome and NETRIN1 Signaling Pathway–Derived Polygenic Risk Scores for Major Depressive Disorder and White Matter Microstructure in the UK Biobank, where they disentangled the specific effects of the NETRIN1 signaling pathway on brain white matter microstructure from general genetic background risk. In a similar vein, Bustillo et al. show in Risk-Conferring Glutamatergic Genes and Brain Glutamate Plus Glutamine in Schizophrenia that a pathway-specific score for glutamatergic neurotransmission correlates directly with observable neurochemical imbalances in the brains of younger patients. By narrowing the focus to specific functional hypotheses, as seen in the language-network analysis by Du et al. in The genetic determinants of language network dysconnectivity in drug-naïve early stage schizophrenia, researchers can drastically reduce statistical noise and map genetic risk directly to specific neural circuits.
Taking this functional refinement further, expression-based polygenic scoring shifts the focus from raw DNA sequences to predicted gene expression levels. In Transcriptome-wide association analysis of brain structures yields insights into pleiotropy with complex neuropsychiatric traits, Zhao et al. demonstrate that aggregating predicted transcriptomic expression levels captures biological signals more effectively than traditional variant-only models, providing a clearer bridge between genotype and phenotype.
Leveraging Shared Architecture and Pleiotropy
Many complex diseases share overlapping genetic architectures, a phenomenon known as pleiotropy. When traits are genetically correlated, multi-trait modeling can leverage the statistical power of one trait to improve risk prediction for another.
In Joint modeling of genetically correlated diseases and functional annotations increases accuracy of polygenic risk prediction, Lu et al. introduce a Bayesian framework that jointly models two correlated diseases alongside functional genomic annotations, significantly outperforming single-trait predictors. This mathematical synergy is further capitalized on in Improving genetic prediction by leveraging genetic correlations among human diseases and traits, where the authors develop a weighted multi-trait framework (wMT-SBLUP). This method approximates complex multivariate modeling by calculating a weighted sum of individual-trait predictors from publicly available summary data, entirely bypassing the need for restricted individual-level genotype datasets.
Clinical Stratification and Risk Forecasting
To transition from a continuous statistical distribution to actionable clinical insights, polygenic scores are used to stratify populations into distinct risk tiers. By identifying individuals at the extreme tails of the distribution, clinicians can uncover hidden vulnerabilities.
As shown by Berndt et al. in Meta-analysis of genome-wide association studies discovers multiple loci for chronic lymphocytic leukemia, aggregating common variants allows for the effective stratification of individuals, revealing that those in the top quintile face nearly double the disease risk of the general population. This stratification is equally valuable for refining risk profiles in patients who already carry high-penetrance mutations. In Evaluation of Polygenic Risk Scores for Breast and Ovarian Cancer Risk Prediction in BRCA1 and BRCA2 Mutation Carriers, Kuchenbaecker et al. demonstrate how polygenic scores can calculate absolute cumulative risks to provide more granular forecasting for mutation carriers. Sue Healey reinforces this in Common variants at 12p11, 12q24, 9p21, 9q31.2 and in ZNF365 are associated with breast cancer risk for BRCA1 and/or BRCA2mutation carriers, demonstrating that while individual common variants have modest effects, their cumulative modeling can estimate significant shifts in absolute disease risk for high-risk cohorts, moving genetics closer to personalized clinical decision-making.
Finally, the clinical utility of these scores follows a distinct temporal pattern. Sun et al. demonstrate in Highly elevated polygenic risk scores are better predictors of myocardial infarction risk early in life than later that highly elevated polygenic scores are far better predictors of disease risk early in life than later. In youth, clinical data is sparse, making genetic risk aggregation uniquely powerful; as patients age, traditional clinical risk factors emerge and capture much of the same variance, positioning polygenic scores as an essential tool for early-life risk forecasting.