Preprint argues LDSC intercept must be fixed to 1 for accurate SNP-heritability estimates

LD Score Regression (LDSC) has become a standard method for estimating whole-genome SNP heritability and genetic correlation from genome-wide association study (GWAS) summary statistics, without requiring access to individual-level genotype data. The approach regresses GWAS test statistics against linkage-disequilibrium scores; its authors previously argued that a free (estimated) intercept in this regression absorbs confounding from population stratification and other biases.

A preprint posted to bioRxiv on 16 May 2026 challenges that assumption directly. The authors argue, using theoretical derivation and simulation, that the free intercept does not accurately capture population-stratification bias and that fixing the intercept to 1 produces more reliable SNP-heritability estimates. If confirmed after peer review, the finding would have implications for a large body of published GWAS-based heritability and genetic-correlation estimates that used the default free-intercept parameterisation.

The preprint has not yet undergone peer review. The authorship and institutional affiliations are not specified in the available lede; readers seeking full methodological detail should consult the bioRxiv record directly. Statistical geneticists and GWAS analysts are the primary audience for evaluating the technical claims; the work is at the methodological research stage and carries no direct clinical implications.