Various genomic and proteomic assays yield high dimensional, irregular functional data. For example, MALDI-MS yields proteomics data consisting of one-dimensional spectra with many peaks. Another example is time course gene expression data from a microarray experiment. In this talk, shall discuss how to use these functional data for cancer classification. I shall propose a unified hierarchical Bayesian model to encompass both the adaptive functional model as well as the classification model. The use of Gibbs sampling with conjugate priors for posterior inference will make the method computationally feasible. I shall compare the performance of the proposed model with other classification methods such as the existing naive plug-in methods by analyzing real cancer related dataset.
Next, I shall demonstrate the use of Bayesian nonparametric model for clustering these functional data. Use of Dirichlet process priors over the basis coefficients will provide a flexible model for functional clustering and can determine the number of clusters automatically. Time allowing, I shall extend this method to cluster gene networks. These methods will be applied to a series of proteomic and genomic data sets.
