Quantile Regression for Distribution Dependent Genetic Effects

13 concepts5 papers1 proposal

Traditional genetic association studies operate under a foundational assumption: a genetic variant shifts the entire phenotypic distribution uniformly. When we run a standard Genome-Wide Association Study (GWAS) or map expression quantitative trait loci (eQTLs), we typically use ordinary least squares (OLS) linear regression. This models how a genetic marker changes the mean of a trait.

However, biology is rarely uniform. A genetic variant might have no detectable effect on the average person, yet act as a powerful driver among individuals who already sit at the extreme high or low ends of a trait's distribution. By focusing exclusively on the mean, traditional models wash out these localized dynamics.

To capture these hidden architectures, researchers have turned to quantile regression. Rather than tracking the mean, this framework slices a phenotype's distribution into distinct percentiles (quantiles), allowing us to uncover distribution-dependent genetic effects.


The Intuition: Heterogeneous Sparsity

Why do mean-based models fail to capture the full picture? The answer lies in a phenomenon known as heterogeneous sparsity.

In a standard linear model, a genetic marker is assumed to have a constant effect size ($\beta$) across the entire population. If you carry the risk allele, your trait value shifts by $\beta$, whether your baseline value was low, medium, or high.

As detailed by Qianchuan He et al. in Regularized quantile regression under heterogeneous sparsity with application to quantitative genetic traits, heterogeneous sparsity means that the genetic architecture itself changes depending on where an individual sits within the phenotypic spectrum. A specific genetic marker might be highly influential for individuals in the 90th percentile of a trait but completely irrelevant for those in the 50th percentile.

Traditional Mean-Based Effect:
Low Quantile  [--- +Beta --->]
Med Quantile  [--- +Beta --->]
High Quantile [--- +Beta --->]  (Uniform shift across the distribution)

Heterogeneous Quantile Effect:
Low Quantile  [ No Effect ]
Med Quantile  [-- +Beta1 -->]
High Quantile [---- +Beta2 ---->] (Effect size and significance vary by quantile)

By utilizing quantile regression, researchers can estimate distinct effect sizes for every single percentile of a distribution. This makes it possible to identify genetic factors that govern complex quantitative traits only under specific physiological conditions or extreme states.


Uncovering Context-Specific Regulation

This distribution-dependent behavior is highly visible in gene regulation. In QRank: a novel quantile regression tool for eQTL discovery, Xiaoyu Song et al. demonstrated that genetic variants often influence gene expression levels at specific quantiles rather than across the entire distribution.

Consider a gene whose expression is tightly controlled by a repressor protein. Under normal conditions (low or average expression), a genetic variant in the promoter region might remain silent. However, when the gene escapes repression and reaches a high expression state (the upper quantiles), that same genetic variant might suddenly become a potent enhancer.

The QRank framework captures these exact dynamics. By testing for associations across multiple quantiles simultaneously, it uncovers regulatory mechanisms that are only active during extreme expression states—insights that traditional mean-based eQTL mapping completely misses.


Moving Beyond the Mean to Map Variance

Quantile-specific effects do more than just pinpoint localized associations; they also explain how genes control phenotypic variability.

If a genetic variant has a positive effect size at the upper quantiles of a trait but a negative or neutral effect size at the lower quantiles, it will naturally stretch the distribution out, increasing its overall spread. Conversely, a variant could act as a buffer, tightening the distribution.

To exploit this property, Jiacheng Miao et al. developed a framework called QUAIL (Quantile Integral Linear Model) in A quantile integral linear model to quantify genetic effects on phenotypic variability. Instead of looking at a single quantile in isolation, QUAIL integrates information across the entire quantile function. This allows the model to robustly identify variance-associated genetic variants (vQTLs). Mapping these variants is crucial for understanding how much an individual's traits might fluctuate in response to environmental changes or therapeutic interventions.


Improving Clinical Prediction and Resolving Missing Heritability

Accounting for distribution-dependent effects has profound implications for disease prediction and population health, particularly for highly skewed metabolic traits.

In Enhancing polygenic risk prediction by modeling quantile-specific genetic effects, Suin Kim et al. introduced the Quantile Regression-based Polygenic Risk Score (QPRS). Traditional polygenic risk scores aggregate mean-effect sizes across thousands of variants, assuming those effects apply equally to everyone. The authors demonstrated that for skewed traits like triglycerides and blood glucose, modeling quantile-specific effect sizes yields significantly more accurate individual risk predictions. If a patient is already predisposed to be in the upper tail of a metabolic trait, a QPRS can evaluate their genetic risk using the weights that specifically matter at that high extreme.

Furthermore, this paradigm helps resolve long-standing statistical discrepancies in genetics. In Genetic associations vary across the spectrum of fasting serum insulin: results from the European IDEFICS/I.Family children’s cohort, Kirsten Mehlig et al. applied quantile regression to childhood fasting insulin levels. The authors discovered that entirely distinct sets of genetic markers were linked to either the low or high extremes of insulin.

Crucially, the authors noted that ignoring these non-linear, quantile-specific architectures causes traditional GWAS to underestimate heritability. By capturing the complex, non-uniform ways that genes interact with phenotypic distributions, quantile-specific modeling bridges the gap between rigid linear assumptions and the nuanced reality of human biology.

Key Foundational Work

Regularized quantile regression under heterogeneous sparsity with application to quantitative genetic traits (2016) — contributed 3 concept elements to this direction.
QRank: a novel quantile regression tool for eQTL discovery (2017) — contributed 2 concept elements to this direction.
History & citationVersion 1 · 2026-10-04

Current · Permanent link

View frozen version 1

Cite this direction

Latest version (a living page):

@misc{lacuna_quantile_specific_genetic_effects_3dda7437,
  title = {quantile-specific genetic effects},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/quantile-specific-genetic-effects/rdr_3dda7437dc65410594f05de5bd0315ad},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-22},
}

This exact version (version 1, frozen text):

@misc{lacuna_quantile_specific_genetic_effects_3dda7437_v1,
  title = {quantile-specific genetic effects},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/quantile-specific-genetic-effects/rdr_3dda7437dc65410594f05de5bd0315ad/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-22},
}
About

These directions overlap with this one through shared papers.

See something wrong? Report a mistake