Statistical Methods fot the Detection of the Parent-of-Origin Effect in Genome-Wide Genotype Data

Genomic imprinting is an epigenetic mechanism that leads to differential contributions of maternal and paternal alleles to offspring gene expression in parent-of-origin (POE) manner. Recently, parent-of-origin effects have attracted attention due to their potential contribution in the explanation of \textquotedblleft missing heritability\textquotedblright. We propose two different procedures for detecting the POEs in genome wide genotype data from related individuals (twins) when the parental origin cannot be inferred. In the first approach we suggest a multistep procedure to detect POEs based on a variance test to evaluate the gene expression heterogeneity between the homozygous and the heterozygous groups. The second method exploits a finite mixture of linear mixed models to propose a test for capturing the presence of POEs; the key idea is that in the case of POEs the population can be clustered in two different groups in which the reference allele is inherited by a different parent. The core advantage of the second method is that it is an integrated procedure developed specifically for the detection of the parental effects, while the first method a multistep procedure which results computationally faster. The performance of the proposed tests are evaluated through a wide simulation study. A discovery analysis on microarray gene expression data of the MuTHER study is performed by both methods.