Genetic Variance Component Analysis Of Human Traits Based On Linear Mixed Effect Models

Variance component analysis has been widely used to analyse the genetic effects on phenotypes.Maternal genetic effects(MGEs)and parent-of-origin effects(POEs)play an important role in genetic mechanism,however,there is no statistical approach specifically tailored for variance component analysis involving MGEs and POEs.Based on mother-child duos data,a new model termed APM model was proposed in this thesis,which orthogonally decomposes phenotypic variances of children into additive,parentof-origin,and maternal genetic effects.This orthogonal decomposition ideally characterizes independent contributions due to each effect,thus is biologically interpretable.Some commonly-used variance component analysis methods were extended to analyze the APM model for both quantitative traits and binary traits.In particular,the restricted phenotype correlation-genotype correlation(R-PCGC)regression method was developed by imposing a non-negative constraint on the eatimation of variance components for case-control studies with binary traits,which yields non-negative and asymptotically unbiased estimates of variance components.Finite sample performances of the considered methods were evaluated through extensive simulations.In particular,the desired properties of R-PCGC(i.e.,asymptotic unbiasedness and non-negativity)were confirmed.Finally,the considered methods were applied to Denmark National Birth Cohort(DNBC)and International Multisite ADHD Genetics(IMAGE)project.Maternal genetic effects were shown to be the major effects for preterm births(46%)in the DNBC study and additive genetic effect were shown to be the major effects(55%)for ADHD in IMAGE study.