Estimating Heritability from Genomic Similarity

18 concepts18 papers1 proposal

For decades, quantitative genetics relied on family and twin studies to estimate heritability—the proportion of phenotypic variance in a population due to genetic differences. For many complex traits, twin studies consistently yielded high heritability estimates, often between 60% and 80%. However, when the era of Genome-Wide Association Studies (GWAS) arrived, researchers faced a paradox: individual genetic variants that passed strict statistical significance thresholds explained only a tiny fraction of this variance.

As noted by Clare Llewellyn et al. in Finding the missing heritability in pediatric obesity: the contribution of genome-wide complex trait analysis, known markers accounted for less than 2% of the variance in children's BMI despite an estimated 80% twin heritability. This discrepancy became known as the "missing heritability" problem. Early GWAS on complex behavioral traits, such as the first childhood anxiety study by Maciej Trzaskowski et al. in Correction: First Genome-Wide Association Study on Anxiety-Related Behaviours in Childhood, frequently returned few or no genome-wide significant hits, confirming that the genetic architecture of these traits is highly polygenic, consisting of thousands of variants with effects too small to detect individually.

The Intuition of Chance Similarity

To capture these elusive small effects, researchers shifted from looking for specific causal variants to estimating the aggregate effect of all common single-nucleotide polymorphisms (SNPs) simultaneously. This methodology, known as Genome-wide Complex Trait Analysis (GCTA) or SNP-based heritability estimation, operates on a simple but powerful intuition: even in a sample of completely "unrelated" individuals, pairs of people will share minute, random fractions of their genomes by chance.

By measuring this background genomic similarity across hundreds of thousands of SNPs and correlating it with phenotypic similarity, we can infer the total additive genetic contribution of common variants. As explained by Nicole Harlaar et al. in Word Reading Fluency: Role of Genome-Wide Single-Nucleotide Polymorphisms in Developmental Stability and Correlations With Print Exposure, this approach acts as a molecular analogue to twin studies. However, by utilizing unrelated individuals, it completely bypasses the environmental confounding and shared-rearing assumptions inherent to family-based designs (Maciej Trzaskowski et al., DNA Evidence for Strong Genome-Wide Pleiotropy of Cognitive and Learning Abilities). If pairs who happen to be slightly more genetically similar also tend to be more phenotypically similar, the trait must be influenced by the measured SNPs.

The Mathematical Engine: Variance Partitioning

How do we mathematically extract this subtle signal from genome-wide data? The standard approach relies on population-level variance component models, specifically linear mixed models (LMMs). Instead of treating each SNP as a fixed effect to be tested individually, the entire genome is modeled as a single random effect.

The total phenotypic variance ($V_P$) is partitioned into additive genetic variance explained by the SNPs ($V_G$) and residual environmental variance ($V_E$). This is typically solved using Restricted Maximum Likelihood (REML) estimation. As demonstrated by Christopher Bryant et al. in Mapping the Genetic Variation of Regional Brain Volumes as Explained by All Common SNPs from the ADNI Study, this framework allows researchers to evaluate the collective contribution of over half a million markers at once.

Crucially, REML can detect a significant genetic signal even when no individual variant achieves genome-wide significance, a phenomenon observed by Jacklyn N. Hellwege et al. in Heritability and genome-wide association study of benign prostatic hyperplasia (BPH) in the eMERGE network and Jianhua Chen et al. in Common variants in ZMIZ1 and near NGF confer risk for primary dysmenorrhoea. Alternatively, Bayesian frameworks can be employed to perform variable selection and model averaging, providing robust heritability estimates that align with prior biological knowledge (Tomi Peltola et al., Bayesian Variable Selection in Searching for Additive and Dominant Effects in Genome-Wide Data).

Narrow-Sense vs. Broad-Sense Heritability

It is important to note that SNP-based heritability provides a lower-bound estimate of true narrow-sense heritability ($h_N^2$), which represents the proportion of variance due to additive genetic effects. Because standard SNP arrays only capture common variants, they miss rare mutations, structural variants, and complex non-additive interactions (such as dominance or epistasis) that contribute to broad-sense heritability ($h_B^2$).

For instance, Essi Viding et al. found in Correction: Genetics of Callous-Unemotional Behavior in Children that while twin studies estimated the heritability of callous-unemotional traits at 64%, SNP-based methods explained almost none of the variance, pointing to rare variants or non-additive effects as the likely source. In contrast, selective breeding studies in other organisms, such as those on Sapium discolor by Yanghui Fang et al. (Joint Selection for Growth and Leaf Color in Superior Trees of Sapium discolor in Fujian Province, China), explicitly calculate the ratio of narrow-sense to broad-sense heritability ($h_N^2/h_B^2$) to determine whether clonal selection or additive breeding strategies will yield the highest genetic gains.

Mapping the Architecture of Complex Traits

SNP-based heritability estimation has fundamentally reshaped our understanding of human traits and diseases by establishing the upper limits of what polygenic scores can realistically predict (Kathryn E. Hatchell et al., Ancestry-specific polygenic scores and SNP heritability of 25(OH)D in African- and European-ancestry populations).

In cognitive science, Maciej Trzaskowski et al. used this approach in DNA evidence for strong genetic stability and increasing heritability of intelligence from age 7 to 12 to prove that the heritability of intelligence increases during childhood while the underlying genetic factors remain stable. Furthermore, bivariate extensions of these models allow researchers to estimate the genetic correlation between different traits. For example, Oliver S. P. Davis et al. demonstrated in The correlation between reading and mathematics ability at age twelve has a substantial genetic component that approximately half of the correlation between reading and math skills is driven by shared common genetic variants, supporting the "generalist genes" hypothesis at the molecular level.

The method also informs clinical research by quantifying the genetic risk of complex conditions. Casey Lynnette Overby, in Estimating heritability of drug-induced liver injury from common variants and implications for future study designs, applied mixed models to show that 30–50% of the risk for drug-induced liver injury is captured by common SNPs, providing a clear justification for scaling up GWAS sample sizes. Similarly, in neuroimaging, Gabriëlla A.M. Blokland et al. (MIR137 polygenic risk for schizophrenia and ephrin-regulated pathway: Role in lateral ventricles and corpus callosum volume) utilized SNP-based heritability to validate structural brain ratios as highly heritable endophenotypes for schizophrenia.

Key Foundational Work

Correction: Genetics of Callous-Unemotional Behavior in Children (2013) — contributed 2 concept elements to this direction.
Correction: First Genome-Wide Association Study on Anxiety-Related Behaviours in Childhood (2013) — contributed 1 concept elements to this direction.
Finding the missing heritability in pediatric obesity: the contribution of genome-wide complex trait analysis (2013) — contributed 1 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_snp_based_heritability_estimation_cde336ca,
  title = {SNP-based heritability estimation},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/snp-based-heritability-estimation/rdr_cde336ca276f42238eaf3758f2b15849},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-19},
}

This exact version (version 1, frozen text):

@misc{lacuna_snp_based_heritability_estimation_cde336ca_v1,
  title = {SNP-based heritability estimation},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/snp-based-heritability-estimation/rdr_cde336ca276f42238eaf3758f2b15849/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-19},
}
About

These directions overlap with this one through shared papers.

See something wrong? Report a mistake