A multi-level model for analyzing whole genome sequencing family data with longitudinal traits

Taoye Chen, Patchara Santawisook, Zheyang Wu

Compared with microarray-based genotyping, next-generation whole genome sequencing (WGS) studies have the strength to provide greater information for the identification of rare variants, which likely account for a significant portion of missing heritability of common human diseases. In WGS, family-based studies are important because they are likely enriched for rare disease variants that segregate with the disease in relatives. We propose a multilevel model to detect disease variants using family-based WGS data with longitudinal measures. This model incorporates the correlation structure from family pedigrees and that from repeated measures. The iterative generalized least squares algorithm was applied to estimation of parameters and test of associations. The model was applied to the data of Genetic Analysis Workshop 18 and compared with existing linear mixed-effect models. The multilevel model shows higher power at practical p-value levels and a better type I error control than linear mixed-effect model. Both multilevel and linear mixed-effect models, which use the longitudinal repeated information, have higher power than the methods that only use data collected at one time point.