Hongyu Zhao PhD
Ira V. Hiscock Professor of Public Health (Biostatistics) and Professor of Genetics and of Statistics
Statistical genomics and proteomics; Bioinformatics; Data integration; High dimensional data; Network and graphical models; Disease risk prediction; Herbal medicine; Microbiome
- Genome Wide Associatio Studies: We are developing statisticla methods to integrate diverse data types and prior biological knowledge to identify genes for common diseases and risk prediction models. The diseases we work on include Crohn's disease, substance abuse, schizophrenia, bipolar, obesity, aneurysm, and autism.
- Network Modeling: We are developing statistical methods to model biological networks under the general framework of Gaussian and other graphical models. Specific networks we are working on include gene expression regulatory networks, signaling networks, and eQTL networks.
- Cancer Genomics: We are developing statistical and computational methods to analyze cancer genomics data, e.g. microarrays and next generation sequencing, to identify cancer subtypes, driver mutations, and appropriate treatments for cancer patients.
- Microbiome: We are developing modeling and analysis approaches for microbiome data generated from next generation sequencing data.
- Proteomics: Our current focus is on targeted proteomics, such as Multiple Reaction Monitoring.
- Herbal Medicine: Through systems biology approach, we are identifying tissue-specific target pathways of herbal medicine.
Our research is driven by the need to analyze and interpret large and complex data sets in biomedical research. For example, in genome wide association studies involving thousands of individuals, millions of DNA variants are analyzed for each person. Such data offer people the opportunity to identify variants affecting disease susceptibility and develop risk prediction models to facilitate disease prevention and treatment. There are many statistical challenges arising from the analysis of such data, including the very high dimensionality, the relatively weak signals, and the need to incorporate prior knowledge and other data sets in analysis. Another example is the analysis of next generation sequence data which present even greater statistical and computational challenges. Our group has been developing statistical methods to address these challenges, such as empirical Bayes methods to borrow information across different data sets, different generalizations of Gaussian graphical models for network inference, Markov random field models for spatial and temporal modeling, and general machine learning methods for high dimensional data.