A0654
Title: Addressing dispersion in mis-measured multivariate binomial outcomes: Analyzing bisulfite sequencing data
Authors: Kaiqiong Zhao - York University (Canada) [presenting]
Abstract: Motivated by a DNA methylation application, the purpose is tackling fitting and inferring a multivariate binomial regression model for outcomes that are contaminated by errors and exhibit extra-parametric variations, also known as dispersion. While dispersion in univariate binomial regression has been extensively studied, addressing dispersion in the context of multivariate outcomes remains a complex and relatively unexplored task. The complexity arises from a noteworthy data characteristic observed in our motivating dataset: non-constant yet correlated dispersion across outcomes. To address this challenge and account for possible measurement error, a novel hierarchical quasi-binomial varying coefficient mixed model is proposed, which enables flexible dispersion patterns through a combination of additive and multiplicative dispersion components. To maximize the Laplace-approximated quasi-likelihood of the model, a specialized two-stage EM algorithm is further developed. Simulations demonstrated that the approach yields accurate inference for smooth covariate effects and exhibits excellent power in detecting non-zero effects. The proposed method is also applied to investigate the association between genome-wide whole blood DNA methylation and levels of ACPA, a preclinical marker for rheumatoid arthritis (RA). The analysis highlights important insights into RA risk factors. The method is implemented in the R Bioconductor package called "SOMNiBUS"
