Research

Our diverse team has expertise in a multitude of areas and has contributed to pushing the boundaries of cancer research forward in critical areas. 

Modern Clinical Trial Design

Clinical trial design has progressed rapidly in the past few decades, leading to a multitude of novel approaches tailored to meet a variety of needs. Specifically, designs have evolved to harness the power of statistics to increase efficiency and minimize the chance of failure. Recognizing this, governing bodies such as the National Cancer Institute (NCI) have encouraged the adoption of such designs. Our group has widespread knowledge in the design and execution of both modern early phase (I-II) and large-scale phase III clinical trials. We have extensive experience with designs such as model based and assisted, seamless phase I-II, and two-stage designs. Our members also work to expand the boundaries of trial design. Dr. Junxiao Hu has developed statistical methodology for group sequential designs with bivariate endpoints and is currently interested in designing phase II trials that simultaneously consider both efficacy and toxicity.  

Developing and Applying Advanced Statistical Methods to Multi-Omics Data

Our group provides consultations for complex omics statistical analyses. We are interested in the development and application of statistical methods for analyzing high-throughput omics data, such as transcriptomics (bulk or single-cell RNAseq), proteomics, and metabolomics data. Although each layer of the omics profile allows a comprehensive view of one particular disease, the development of a disease often depends on cross-talk between multiple molecular layers. This requires multi-omics data integration and examination of how key interactions between gene transcripts and/or proteins contribute to human diseases. Recently, Dr. Yonghua Zhuang and his colleagues have developed several statistical methods, including a novel tissue augmented Bayesian model for eQTL analysis, sparse multiple canonical correlation network analysis (smCCNet), and an augmented high-dimensional graphical Lasso method to incorporate prior biological knowledge for global network learning (ahGlasso). He also developed Graph Convolutional Neural Network models for multi-omics classification of COPD.

Head and Neck SPORE

Investigators at the University of Colorado Cancer Center successfully obtained funding for a Specialized Program of Research Excellence (SPORE) from the National Cancer Institute to study head and neck squamous cell carcinoma (HNSCC). The SPORE’s research is divided among three projects, which aim to 1) explore how blocking certain interactions between tumor and immune cells may enhance the effect of radiation therapy (RT), 2) determine whether combining immunotherapy with RT can overcome resistance to RT, and 3) evaluate the potential of a novel drug that inhibits protein synthesis. The goal is to translate the findings from these projects into new therapeutic strategies that will improve survival and quality of life for HNSCC patients.

Junxiao Hu, Yonghua Zhuang, Dexiang Gao, Kathleen Torkko, and Andrew Nicklawsky provide the SPORE with essential biostatistics support through the Data Science Core (DSC), directed by Dexiang Gao. The DSC assists with several facets of SPORE research: hypothesis formulation; clinical trial design; sample size determination; analytical and database design; and data collection, analysis, management, and storage. Currently, the DSC is developing a pipeline and a user-friendly R Shiny app to help investigators analyze variant allele frequencies.

More information about the Colorado Head and Neck Cancer SPORE can be found here.

R Package Development for Improved Quality and Efficiency

Alyse Staley and Vida Alami of the Cancer Center BBSR Biostatistics Core are developing an R package to improve internal efficiency through automated functions and standardized templates. Before delving into data analysis, it is crucial to ensure that the data are high quality by inspecting the univariate distributions of each variable, the bivariate relationships between variables, and the patterns of missing data. Our R package aims to make this routine data checking process easier. It will include functions to help quickly clean and format data, as well as a template that automatically generates figures to examine the patterns of interest. Once the data are clean, our package will streamline analysis by providing functions to create standardized tables, commonly used plots, and already-formatted HTML and Word reports. By automating the tedious, but necessary, steps of data projects, we can focus on our favorite aspect of the job—analyzing and interpreting the data! Check out our “Improving Efficiency with R” lecture to learn more about implementing efficient practices for your data projects.

Animal Studies

In addition to clinical trials and big data studies, the Core can assist researchers with their animal experiments. Our expertise includes assisting with experimental design (e.g., identifying the correct experimental unit, addressing randomization and blinding), calculating appropriate sample sizes, writing statistical plans and power and sample size sections for grants and papers, and analyzing the results in view of the special statistical issues involving small sample sizes (n<10) that are often the case with animal studies. Our goal is to help plan well-designed experiments to ensure that only the minimum number of animals necessary are used. Dr. Kathleen Torkko has worked with numerous scientists conducting animal research and was the Course Director for BIOS6606, the statistics course that focused on basic science research and was required for Graduate School students.