A dissertation defense by Xiangning Xue, Department of Biostatistics, School of Public Health.

Committee:
George Tseng, Biostatistics (advisor, chair)
 

Abstract:
Circadian rhythms are endogenous processes that synchronize biological activities to a 24-hour day
light cycle. Circadian rhythms exhibit significant variations across individuals, influenced by factors such
as sex and age, and are also found to be linked to health status, particularly in psychological diseases.
Circadian analysis is performed to identify differential rhythmicity like oscillating amplitude and phase
between healthy controls and subjects with diseases to explore the underlying molecular mechanism
and etiology of dysfunctions.

In this thesis, the first paper presents a comprehensive and interactive pipeline to capture the
multifaceted characteristics of differentially rhythmic biomarkers with accurately controlled type I error.
Analysis outputs are accompanied by informative visualization and interactive exploration. The
workflow is demonstrated in a real world study and is extensible to general omics circadian
applications.

In human studies, the internal circadian clock of a subject may be different from recorded time due to
measurement error or subject-specific variability. To accommodate this variability, the second paper
develops a unified Bayesian model to simultaneously re-estimate molecular circadian time of each
subject and detect circadian genes.

In the third paper, we extend the Bayesian circadian model to integrate multiple cohorts (e.g., multiple
brain regions). Biomarker detection by Bayes factor allows integration of homogeneous information
across cohorts while distinguishes heterogeneous circadian signals in individual cohort.

Public health significance: Brain’s circadian mechanism plays a critical role in aging as well as
many psychiatric disorders. The proposed methodologies provide statistical and bioinformatic tools for
disease understanding and identification of potential drug targets.