Biologically Informed Weighting for Polygenic Risk Scores

This direction is split into other directions: functional genomic prioritization. Its last version stays readable and citable here.

Version 1 of 173 papers110 conceptsfrozen 2026-10-04

Traditional genetic risk models often struggle with a "signal-to-noise" problem. While Genome-Wide Association Studies (GWAS) identify thousands of genetic variants associated with diseases, most of these variants are "passengers"—markers that appear significant only because they are physically close to a true causal mutation, a phenomenon known as linkage disequilibrium.

To solve this, researchers have moved toward biologically informed weighting, a strategy that uses functional genomic maps to prioritize genetic variants located in active regulatory regions. By treating the genome not as a flat string of letters but as a structured landscape of "switches" and "coding zones," these models can distinguish causal signals from statistical noise.

The Problem with Statistical Agnosticism

Standard Polygenic Risk Scores (PRS) are typically "agnostic." They treat every single-nucleotide polymorphism (SNP) as equally likely to be functional, weighting them solely based on their statistical association in a specific study. As noted by Quanfa He and colleagues, this approach ignores the biological context of mutations, such as whether they occur in regulatory regions like enhancers or promoters.

This lack of context creates two major hurdles:

  1. Linkage Disequilibrium (LD) Ambiguity: In regions where many variants are inherited together, statistics alone cannot tell which one actually disrupts a biological process.
  2. Sample Size Bottlenecks: Small studies often lack the power to reach strict significance thresholds, leaving many true risk variants "buried" in the noise.

Functional Annotations as Biological Filters

The core intuition behind the new generation of models is to use functional annotations—external datasets that map where genes are turned on, where proteins bind to DNA, and which regions are evolutionarily conserved—as a "prior" probability.

Frameworks like AnnoPred, developed by Qiongshi Lu and Yiming Hu, use a Bayesian framework to assign higher weights to variants located in these biologically active regions. In Leveraging Functional Annotations in Genetic Risk Prediction for Human Complex Diseases, the authors demonstrate that by prioritizing variants in tissue-specific regulatory regions, they can significantly improve the accuracy of disease risk prediction compared to methods that treat all genomic regions equally.

Mechanisms of Prioritization

Researchers use several distinct mechanisms to integrate this biological "prior knowledge" into statistical models:

  • Bayesian Upweighting: Models like GenoWAP and LSMM (Jingsi Ming et al.) treat GWAS p-values as a distribution that can be "rectified" by functional scores. If a variant has a moderate statistical signal but sits in a known brain-specific enhancer, the model "rescues" it by increasing its weight.
  • Tissue-Specific Mapping: Tools like GenoSkyline-Plus (Qiongshi Lu et al.) allow researchers to focus only on the "switches" active in relevant tissues. For example, Systematic tissue-specific functional annotation of the human genome revealed that genetic risk for Alzheimer’s is more concentrated in immune cell enhancers (monocytes) than in general brain tissue.
  • Kernel-Based Weighting: In rare-variant studies, researchers like Remo Monti and Stefan Konigorski use kernels to group mutations by shared biological roles. Instead of counting every mutation as "1," they weight them by predicted deleterious impact on protein stability or splicing.

Why It Works: Reducing the Search Space

Biologically informed weighting works because it effectively reduces the "search space" for causal variants. By focusing on the ~10% of the genome that is functionally active in a given context, models can ignore the 90% that acts as background noise.

This has immediate practical benefits. In studies of ADHD, Quanfa He found that functionally-annotated polygenic scores were more effective at predicting inattention symptoms in adolescents than traditional methods. Similarly, the DeepGWAS framework (Yun Li et al.) showed that integrating brain-specific annotations can increase the discovery power of a study by up to 3X without needing a single additional participant.

Beyond Linear DNA

The most recent advancements extend this intuition into three dimensions. The PSCAN method (Zheng-Zheng Tang et al.) prioritizes variants that cluster together in the 3D folded structure of a protein, even if they are far apart on the linear DNA sequence. This recognizes that the "functional context" of a mutation is defined by its physical environment in the cell, not just its coordinate on a chromosome.

By shifting from agnostic statistical testing toward biologically informed modeling, this research direction provides a clearer path to understanding how the "dark matter" of our non-coding DNA actually drives human health and disease.

Papers in this version

Showing 60 of 73 papers.

History & citationVersion 1 · 2026-10-04

Split into other directions · Permanent link

View frozen version 1

Cite this direction

Latest version (a living page):

@misc{lacuna_biologically_informed_weighting_for_poly_3268ddd9,
  title = {Biologically Informed Weighting for Polygenic Risk Scores},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/biologically-informed-weighting-for-polygenic-risk-scores/rdr_3268ddd94c5441be808b401bd32791c5},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-20},
}

This exact version (version 1, frozen text):

@misc{lacuna_biologically_informed_weighting_for_poly_3268ddd9_v1,
  title = {Biologically Informed Weighting for Polygenic Risk Scores},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/biologically-informed-weighting-for-polygenic-risk-scores/rdr_3268ddd94c5441be808b401bd32791c5/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-20},
}