Dr. Michele Guindani
The University of Texas MD Anderson Cancer Center
Department of Biostatistics
The core of my research is to develop novel statistical methods for the analysis of heterogeneity in complex datasets, and for the purpose of data integration. Identification of relevant subtypes of the population of interest is of particular importance in many contexts. For example, tumor subtypes based on patterns of gene expression or aberration content have shown moderate association with clinical outcomes, expressed in terms of disease-specific survival duration and recurrence (e.g. Chin et al, 2006). Subsets of genes found to be associated with a poor outcome are particularly interesting as therapeutic targets for treatment of patients that are refractory to current therapies. Understanding the associations and patterns within a dataset is even more important with the explosive growth of data acquisition technologies experienced in modern science. The effortless access to a wide array of different data platforms makes it relevant to integrate several sources of information and aim at a global approach to inference.
By using innovative Bayesian nonparametric (NP) techniques, we can identify hidden clusters and subtypes in the data, as well as provide a coherent probabilistic framework for data integration. Some examples of the type of research I am interested in include:
- develop novel statistical methods to estimate progression-free survival in prostate cancer using longitudinal information on biomarkers associated with a therapeutic outcome.
- develop novel statistical methods to identify tumor subtypes on the basis of gene expression measurement and other genetic information.
- develop novel statistical methods for imaging genetics: integration of brain imaging data and genetic information.
- develop a decision theoretic framework and novel procedures for control- ling the number of false positives in general decision problems.
- be involved in translational research collaborations with other investigators in the UT Health Science Center in order to motivate and guide new applications of the methods developed under the previous aims.
- develop code and user-friendly software for the dissemination of the pro- posed methods.
Keywords: Statistical theory and methods: Hierarchical functional data analysis; Multivariate nonlinear statistical methods for classification and prediction problems; Spatial data analysis; Hierarchical Bayesian modeling and computation; Markov Chain Monte Carlo algorithms; Semiparametric/ Nonparametric methods and mixed models.
Applications: High-throughput functional genomics experiments in nutrition and cancer; Complex multivariate datasets in biostatistical and bioinformatics applications; Dose response modeling and synergy assessment; Environmental applications; Risk modeling. A tutorial in our laboratory would provide advanced training in state of the art statistical methods, advanced mathematical modeling and computer programming and software development.
Office: MDA FCT 4.6024 (Unit 1411)
Title: Assistant Professor
Ph.D. - Universita' Bocconi - 2005