Properties and Evaluation of the MOBIT – a novel Linkage-based Test Statistic and Quantification Method for Imprinting

Genomic imprinting is a parent-of-origin effect apparent in an appreciable number of human diseases. We have proposed the new imprinting test statistic MOBIT, which is based on MOD score analysis. We were interested in the properties of the MOBIT concerning its distribution under three hypotheses: (1) H0, a: no linkage, no imprinting; (2) H0, b: linkage, no imprinting; (3) H1: linkage and imprinting. More specifically, we assessed the confounding between imprinting and sex-specific recombination frequencies, which presents a major difficulty in linkage-based testing for imprinting, and evaluated the power of the test. To this end, we have performed a linkage simulation study of affected sib-pairs and a three-generation pedigree with two trait models, many two- and multipoint marker scenarios, three genetic map ratios, two sample sizes, and five imprinting degrees. We also investigated the ability of the MOBIT to quantify the degree of imprinting and applied the MOBIT using a real data example on house dust mite allergy. We further proposed and evaluated two approaches to obtain empiric p values for the MOBIT. Our results showed that twopoint analyses assuming a sex-averaged marker map led to an inflated type I error due to confounding, especially for a larger marker-trait locus distance. When the correct sex-specific marker map was assumed, twopoint analyses have a reduced power to detect imprinting, compared to sex-averaged analyses with an appropriate correction for the inflation of the test statistic. However, confounding was not an issue in multipoint analysis unless the map ratio was extreme and marker spacing was sparse. With multipoint analysis, power as well as the ability to quantify the imprinting degree were almost equally high when a sex-averaged or the correct sex-specific map was used in the analysis. We recommend to obtain empiric p values for the MOBIT using genotype simulations based on the best-fitting nonimprinting model of the real dataset analysis. In addition, an implementation of a method based on the permutation of parental sexes is also available. In summary, we propose to perform multipoint analyses using densely spaced markers to efficiently discover new imprinted loci and to reliably quantify the degree of imprinting.