Imputing the parental origin of the sibling's haplotype from parental phenotypes

To recruit cases for late-onset disease study is challenging since these diseases occur in elder people. Moreover, typically we have a very limited number of late-onset disease cases in Biobank data. But, on the other hand, the parental disease status may be available by questionnaire. Because of this, methods have been developed to utilize parental disease status instead Liu et al. (2017); Hujoel et al. (2020). In these approaches, the late-onset phenotype of the participant is imputed from parental statuses. And, downstream, a genome-wide association study (GWAS) is performed using the participants genotype and imputed phenotype. In this paper, we take another view on utilizing parental phenotypes. We treat this problem as missing parental genotype rather than missing participants phenotype. First, we propose an imputation scheme to infer the parental origin of the participants genotype from a collection of extra parental phenotypes (non-focal phenotypes) and the participants genotype. Second, we propose a computationally efficient approach to incorporate the imputed parental origin information into the downstream GWAS. We explore the feasibility of the proposed two-step approach on simulated and real data. And we derive the power increase of GWAS as a function of imputation quality. These results indicate that the imputation scheme needs about 100 non-focal phenotypes to achieve enough accuracy to facilitate the GWAS downstream.