statistical genetics · BEGINNER LESSON
LDSC
LD score regression (LDSC) is a statistical method that estimates heritability, and sometimes genetic correlation, from genome-wide association study (GWAS) summary statistics plus linkage disequilibrium (LD) scores. It uses how association test statistics change with LD to infer how much trait variation is linked to genetic variation.
What LDSC is and what it measures
LD score regression (LDSC) is a method for analyzing genome-wide association study (GWAS) summary statistics, meaning the association results reported for many genetic markers without needing each person’s full genotype data. It estimates heritability, which here means the proportion of trait variation explained by genetic effects, and it can also be used for genetic correlation, which describes shared genetic architecture between two traits. The main input is the GWAS test statistic for each single-nucleotide polymorphism (SNP), a SNP being a single DNA position that varies across people. LDSC also uses an LD score for each SNP, where linkage disequilibrium (LD) means nearby genetic variants are correlated and the score summarizes that correlation pattern. [0,1,2]
[1] Discovering non-additive heritability using additive GWAS ... [2] Estimation of Genetic Correlation via Linkage Disequilibrium Score ... [3] Leveraging LD eigenvalue regression to improve the estimation of ...
How it works
The core idea is that, for polygenic traits, SNPs in higher LD with their neighbors tend to have larger test statistics on average because they tag more causal variants. LDSC regresses SNP association statistics on LD scores to estimate how much of the observed signal fits a polygenic pattern rather than simple noise. In the heritability setting, the target is the additive contribution of genotyped SNPs. In the genetic-correlation setting, LDSC uses summary statistics for two traits and the LD score structure to estimate their shared additive genetic covariance. [0,1,2]
[1] Discovering non-additive heritability using additive GWAS ... [2] Estimation of Genetic Correlation via Linkage Disequilibrium Score ... [3] Leveraging LD eigenvalue regression to improve the estimation of ...
Limitations
LDSC is computationally efficient because it uses summary statistics, but that speed comes with tradeoffs. The supplied sources note that standard errors can be large, especially when few SNPs are available or when LD scores come from a reference sample that is not well matched to the GWAS sample’s ancestry. The method also relies on assumptions of an approximately polygenic architecture; if those assumptions are badly violated, estimates can be affected. One source also notes that LDSC mainly uses diagonal parts of the LD matrix, so it does not fully use all LD information. [0,1,2]
[1] Discovering non-additive heritability using additive GWAS ... [2] Estimation of Genetic Correlation via Linkage Disequilibrium Score ... [3] Leveraging LD eigenvalue regression to improve the estimation of ...
SEE THE IDEA
How LDSC connects GWAS data to heritability
Association test results for SNPs, without individual-level genotypes
How strongly each SNP is correlated with nearby SNPs
Relates test statistic size to LD score
Heritability or genetic correlation
Higher LD scores can be associated with larger test statistics, and LDSC uses that pattern to infer genetic parameters.
LDSC summarizes a genome-wide pattern in association statistics using LD information, then converts that pattern into a genetic estimate.
TRY IT YOURSELF
Test your understanding
Choose an answer, then check the explanation.
PUT IT TOGETHER
Match each concept to its meaning
Drag a label onto its description, or select a label and then select a description. This also works with a keyboard.
4 pairs to match
Sources and further reading
This general lesson was generated from public source material. It explains a concept; it does not interpret a private paper or provide medical advice.