Limitations of GCTA as a solution to the missing heritability problem

Area of Science:

• Genetics
• Statistical Genetics
• Bioinformatics

Background:

• Genome-wide association studies (GWASs) analyze relationships between phenotypes and single-nucleotide polymorphisms (SNPs).
• Genome-wide complex trait analysis (GCTA) emerged to capture heritability missed by early GWAS analyses.
• GCTA has been applied to diverse complex traits, but its heritability estimates are increasingly questioned.

Purpose of the Study:

• To evaluate the reliability and stability of heritability estimates derived from GCTA using current SNP data.
• To identify the factors contributing to potential biases and inaccuracies in GCTA heritability estimates.

Main Methods:

• Analysis of the sensitivity of GCTA to singular values of the genetic relatedness matrix (GRM).
• Mathematical proof demonstrating bias in heritability estimates when GCTA assumptions are met.
• Investigation of GCTA performance with stratified populations and skewed GRM singular values.
• Assessment of GCTA's sensitivity to sample selection and phenotypic measurement error.

Main Results:

• GCTA's heritability estimates are highly sensitive to all singular values of the GRM.
• When GCTA assumptions hold, estimates are biased and standard errors are inaccurate.
• Stratified populations lead to skewed GRM singular values, causing overfitting and inflated heritability.
• Estimates vary significantly based on sample choice and phenotype measurement accuracy.

Conclusions:

• GCTA cannot produce reliable or stable heritability estimates from current SNP data.
• The method is prone to overfitting due to issues with estimating small singular values in the GRM.
• Results from GCTA require cautious interpretation, especially concerning their qualitative implications.