More specifically, our research focuses on:

High-throughput single-cell sequencing and pathway modeling of cancer heterogeneity

Cancer is a leading cause of death worldwide. Though many drug treatments, including chemotherapy, had achieved great successes in reduction of cancer mortality, drug resistance has long been identified as a major reason for therapy failure in cancer patients. Tumor heterogeneity, the existence of subpopulations of cells in cancers is considered to influence the sensitivity of the tumor cells to drug treatment. Thus, cancer therapy needs to become more personalized, selective, and specific. Single-cell transcriptome analysis by RNA-sequencing (RNA-seq), which assesses gene expression abundance at single cell resolution, provides great opportunities to analyze intratumoral heterogeneity systematically. We aim to develop a series of statistical and computationally efficient methods on both intratumoral heterogeneity clustering and significant pathway identification, which could be used for cancer subtype identification and treatment optimization.

Integrative learning for disease biomarker discovery and network analysis

Biomarkers are widely used in diagnosing diseases, monitoring treatments, and evaluating potential drug candidates. Integrating the knowledge from multi–omic domains accelerates the advancements of biomarker discovery. Combining multi-omic data (e.g. genome, transcriptome, epigenome, proteome, metagenome, metabolome and clincal variables) of the same patient cohort enables us to more accurately investigate disease subtypes, detect disease associated or driver genes and look into related regulatory network. The ultimate goal here is translational "precision medicine" to better diagnose and treat patients.

Machine learning approach on multi - omics data for human oral disease

Recent studies have provided numerous evidences of linking oral microbiome to periodontal disease and caries and other diseases. Current oral microbiome research has generated large amounts of –omics data such as metagenomics, metatranscriptomics, metaproteomics along with other types of data such as environmental factors, host genomic and clinical data. These multi-omics data provide snapshots of molecular activities of oral microbiome; however, there is a lack of methods for the integration analysis at the system level. The Human Microbiome Project provides a census of oral microbiome in healthy individuals. Most research involving diseased individuals has a small number of subjects. In this research, we aim to develop statistical analysis tools to integrate these publicly available multiple types of oral microbiome data from multiple studies for better understanding the initiation and progression of periodontal disease and caries in adults and children, and for accurate prediction of disease status via machine learning approaches. The knowledge and integrative methods gained from this project can be extended to other diseases.

Tracing microbial evidence in forensic studies

With the advances in genomic sequencing technologies, there is an increased interest in studying the forensic potential of microorganisms, in particular when there is an absence of probative human DNA. Materials collected at crime scenes could be connected to individuals through information the microbial community reveals. A gap exists in the availability of well-designed computational methods to analyze the microbiome data, particularly for metagenomic identification of the host. Laboratory-based method development can suffer greatly in the absence of a well-organized statistically sound method to evaluate such large and diverse data. We aim to address this need through the development of computational methods to evaluate human microbiome as source of forensic evidence.

Metagenomic analysis and applications in biological threat detection

High-throughput next generation sequencing technologies provide a powerful way to detect biological threats from metagenomic samples taken directly from the environment without prior knowledge of sample composition. In the analysis of metagenomic, more specific, metatranscriptomic, data sets, we can examine and compare the active gene functions and pathways in the environmental or host-associated metagenomic samples with the presence or absence of biological threat agents (organisms or viruses). This is accomplished by identifying which genes are active in a sample and characterizing which functional patterns are associated with the presence of biothreat agents. Moreover, functional analysis of metagenomes can explore how functional diversity of microbial communities correlate with important biological factors of interest including the presence of a particular threat organism and its virulence level. In this research we propose to build rigorous statistical models and rapid computational algorithms to detect biological threats based on metagenomic sequencing data.