Statistical methods for genetic and bioinformatic studies

点击翻译按钮获取中文摘要

The University of Guelph is well-known for its longstanding tradition of excellence in agricultural, animal, biological, and food science. Statistical and computational tools are indispensable for research in these disciplines.  My long-term goal is to develop a statistical and computational lab on campus to address issues and analyze data arising from cutting-edge research in these areas. Specifically, my research program mainly comprises two themes: statistical bioinformatics and statistical genetics. The next generation sequencing (NGS) technique enables the generation of  a massive amount of metagenomic data for the exploration and detection of relationships between biological/environmental factors and microbiome composition, and thus their impacts on human health. Research projects in the first theme, statistical bioinformatics, of my research program include: 1) model-based clustering of human gut bacterial microbiome communities and the association between confounding covariates and the underlying cluster structure.2) bacterial association analysis and detecting differential abundant bacteria based on count data and compositional data; and 3) phage virome sequencing data preprocessing pipeline development and statistical methodology development for downstream analysis. Our research will generate useful statistical and computational tools for analyzing the high throughput bacterial and phage "omic" data separately or integratedly.  In statistical genetics, my research interests focus on statistical methodology and software tool development for single nucleotide polymorphism (SNP) data analysis. My research projects include: 1) genome-wide association analysis using longitudinal data with missing phenotype data; 2) detecting gene-environmental interactions and time-varying genes via longitudinal studies; and 3) optimal design for animal selection and animal mating using genomic information. In 1), we study the impact of missing data on the parameter estimates and, subsequently, their impacts on genetic association tests. We propose a likelihood-based approach via the Monte-Carlo expectation-maximization (MCEM) algorithm for model fitting that leads to a robust downstream association analysis. In 2), we propose to detect the gene-environmental interactions of time-varying genes by introducing a random slope in a mixed effects model to capture the subject-specific genetic interaction effect with a covariate when repeated measurements within subjects allow for the prediction of the random slope.  In 3), recent technology allows breeders to select animals based on genomic evaluation using a high-density SNP panel (60K). The next step is assigning animals for mating to maximize their expected progenies' genomic breeding values of important economic traits subject to some constraints. Our research will focus on optimal strategies for animal selection and animal mating in different types of herds and stages.