Statistical Models for Detecting Transgenerational Genetic Effects
Genome-wide association studies (GWAS) have successfully discovered a number of genes that control disease susceptibility and variation in quantitative traits. Despite the large number of genes found to be associated with human diseases and complex traits, a limited amount of the total heritability is explained by these discoveries. One hypothesis is that some of this missing heritability is due to transgenerational effects, effects of genetic factors in one generation that affect the phenotypes in a subsequent generation without Mendelian transmission of alleles.
The ability to detect transgenerational effects in humans is mainly limited to maternal effects when using epidemiological data. Furthermore, currently available methodologies lack approaches to identify associations between maternal-fetal genotype (MFG) interactions and quantitative traits for arbitrary family structures. To address this issue, I present the Quantitative-MFG (QMFG) test, a linear mixed effect model in which maternal and offspring genotypes are considered fixed effects and residual familial correlations are random effects. This approach handles pedigrees of virtually any size, common or unusual scenarios of MFG incompatibility, and additional covariates. Another attractive feature of the QMFG test is the ability to easily extend the approach to multiple loci. With simulation studies, I demonstrate the statistical validity of the QMFG analysis method and show that if a standard model, which considers only offspring genotypes, is fit to data with an MFG effect, associations can be missed or misattributed. To allow other researchers to determine if there is evidence of MFG effects in their own data, I have developed and implemented subroutines as part of the software program Mendel, which is freely available. The QMFG test may provide another approach to uncovering sources of missing heritability in association studies.