For decades, medical genetics focused on "monogenic" diseases—conditions like cystic fibrosis or Huntington’s disease caused by a single, high-impact mutation. However, most common human traits and diseases, from height to schizophrenia, do not follow this simple pattern. Instead, they are "polygenic," shaped by the cumulative influence of thousands of tiny genetic variants across the entire genome.
Polygenic risk scoring (PRS) is the mathematical framework used to aggregate these microscopic effects into a single, continuous metric of genetic liability. By summing the weighted presence of risk alleles identified in large-scale genome-wide association studies (GWAS), researchers can transform high-dimensional genomic data into a predictive tool for clinical risk.
Aggregating the Microscopic
The intuition behind a polygenic score is that while a single genetic variant might increase the risk of a condition by only 0.01%, the collective burden of thousands of such variants can be substantial. As Robert Maier and Naomi R. Wray argue in Embracing polygenicity: a review of methods and tools for psychiatric genetics research, the field has shifted from searching for "risk genes" to quantifying "genetic liability."
In practice, a PRS is calculated by taking an individual's genotype at millions of positions (SNPs) and multiplying the number of risk alleles they carry by the effect size of each allele. This summation acts as a "genetic blueprint" or a proxy for a trait. For example, Paroma Varma demonstrated that polygenic scores for aortic valve area can predict future heart disease even in individuals who have not yet undergone cardiac imaging.
Refining the Signal with Bayesian Modeling
A major challenge in polygenic scoring is "linkage disequilibrium"—the tendency for neighboring genetic variants to be inherited together. If two variants are correlated, simply summing their effects would lead to double-counting.
To solve this, researchers use advanced statistical frameworks like LDpred2, developed by Florian Privé and colleagues in LDpred2: better, faster, stronger. This method uses a Bayesian approach to model the distribution of effect sizes across the entire genome simultaneously. By applying "shrinkage" to the weights of variants in dense genomic regions, these models account for correlations and improve the accuracy of risk predictions, especially when dealing with complex traits like blood lipids or psychiatric disorders.
The Power of the Tail
While a polygenic score provides a value for everyone, its greatest clinical utility lies at the extremes of the distribution. Individuals in the "top tail"—typically the highest 2% of genetic risk—often face a disease likelihood equivalent to those carrying rare, high-impact monogenic mutations.
In Deep-coverage whole genome sequences and blood lipids among 16,324 individuals, Pradeep Natarajan showed that a high polygenic score is ten times more prevalent than monogenic mutations in people with extremely high LDL cholesterol. Similarly, Atlas Khan found that individuals in the top 2% of the polygenic distribution for chronic kidney disease face a three-fold higher risk, mirroring the risk profile of someone with a strong family history.
Transdiagnostic and Multi-Trait Integration
The genetic architecture of many conditions is not isolated; many genes influence multiple traits simultaneously, a phenomenon known as pleiotropy. Modern research leverages this by creating "transdiagnostic" or "multi-polygenic" scores.
For instance, Alexander Neumann showed in Combined polygenic risk scores of different psychiatric traits predict general and specific psychopathology in childhood that combining scores for various adult psychiatric disorders significantly improves the prediction of mental health struggles in children. This suggests a "general psychopathology factor" (or p-factor) where shared genetic liability drives a broad range of symptoms. James J. Li further refined this in RDoC Mechanisms of Transdiagnostic Polygenic Risk for Trajectories of Depression, demonstrating how these shared genetic risks specifically predict worsening depression trajectories from adolescence into adulthood.
Portability and Ancestry
A significant hurdle for the field is that most GWAS data comes from individuals of European descent, making scores less accurate for other populations. Lars G. Fritsche and Daiwei Zhang have addressed this in On cross-ancestry cancer polygenic risk scores by developing recalibration techniques. By adjusting scores to ancestry-specific distributions, researchers can still identify high-risk individuals in diverse populations, ensuring that the benefits of precision medicine are more equitably distributed.