Background and goal
The epigenetic phenomenon of imprinting is critical to understanding disease expression. However, it is hard to detect and can be species- and tissue-specific. One approach is to detect imprinting by taking advantage of linkage information. Although imprinting detection methods exist, the effects of potential confounders, such as heterogeneity, sex-specific penetrance, and differential sex-based ascertainment, have not been explored in depth. In this study we explored possible confounders using two different imprinting detection approaches. Our goal was to understand the essence of how imprinting and linkage interact and to elucidate the underlying issues in existing imprinting detection approaches.
One method (PP) models imprinting by maximizing lod scores with respect to parent-specific penetrances. The other method (DRF) approximates imprinting by maximizing two-point lods with respect to differential male-female recombination fractions. We compared power, type 1 error, and confounder effects in these two linkage-based imprinting detection methods using two-point linkage analysis for simplicity. We computer-simulated data, determining power and type 1 error for imprinting detection among datasets with detectable linkage. We generated data with and without imprinting, with and without heterogeneity, and with varying reduced penetrance, family and dataset size. We also examined non-imprinting situations that could mimic imprinting, e.g., sex-specific penetrances, and a scenario requiring a sex-specified affected parent for ascertainment.
Without heterogeneity, PP had more imprinting-detecting power than DRF. Surprisingly, PP’s power increased when parental affectedness status was ignored, but decreased with heterogeneity. With heterogeneity, type 1 error could increase dramatically for both methods. However, DRF’s power also appeared to increase under heterogeneity, more than could be attributed to the inflated type 1 error. We determined the reasons behind these phenomena.
The presence of sex-specific penetrance increased false positives for PP but not for DRF. Ascertainment through an affected “mother” in unimprinted data did not lead to false positives with either method. For PP, increased information may depend on non-penetrant heterozygous individuals, arguing against using affected sib pairs or other affecteds-only methods.
The high type 1 error levels under some circumstances means methods must be used cautiously. Using differential recombination fractions to approximate imprinting has certain advantages and should be incorporated into future imprinting detection programs.