Kinship estimation bias carries over to heritability estimation bias using variance components

Area of Science:

• Population Genetics
• Quantitative Genetics
• Biostatistics

Background:

• Heritability quantifies genetic contributions to traits, crucial for disease research.
• Variance components models require kinship matrices (Genetic Relatedness Matrices/GRMs) for heritability estimation.
• Standard kinship estimators are known to be biased in structured populations.

Purpose of the Study:

• To characterize heritability estimation biases caused by kinship estimation in structured populations.
• To evaluate the performance of different kinship estimators under population structure.
• To identify an unbiased method for heritability estimation in diverse cohorts.

Main Methods:

• Derived a closed-form expression for heritability bias in the ratio-of-means (ROM) estimator.
• Utilized simulation studies with admixture and family structures using 1000 Genomes genotypes.
• Analyzed three real-world structured datasets with phenotypes: San Antonio Family Study, HCHS/SOL, and a nephrotic syndrome cohort.

Main Results:

• The standard mean-of-ratios (MOR) estimator shows significant bias, more so than ROM, due to low-frequency variants.
• Popkin, an unbiased population kinship estimator, produced accurate heritability estimates in structured settings.
• Pedigree-only estimates exhibited upward heritability bias in structured populations.
• The MOR estimator demonstrated variable biases (downward/upward) based on population structure and variant frequency.

Conclusions:

• Standard kinship estimators (ROM, MOR) lead to biased heritability estimates in structured populations.
• The Popkin estimator overcomes biases associated with population structure.
• Accurate heritability estimation in diverse populations necessitates the use of unbiased kinship estimators like Popkin.