Nonparametric Bayes Differential Analysis of Multigroup DNA Methylation Data
share
DNA methylation datasets in cancer studies are comprised of sample measurements on a large number of genomic locations called cytosine-phosphate-guanine (CpG) sites with complex correlation structures. A fundamental goal of these investigations is the development of statistical techniques that identify disease genomic signatures across multiple patient groups determined by different experimental or biological conditions. We propose BayesDiff, a nonparametric Bayesian approach for differential analysis relying on a novel class of first order mixture models called the Sticky Poisson-Dirichlet process or two-restaurant two-cuisine franchise (2R2CF). The BayesDiff methodology flexibly utilizes information from all CpG sites, adaptively accommodating any serial dependence in the data corresponding to the widely varying inter-probe distances, to perform simultaneous inferences about the differential genomic signature of the patient groups. In simulation studies, we demonstrate the effectiveness of the BayesDiff procedure relative to existing statistical techniques for differential DNA methylation. The methodology is applied to analyze a gastrointestinal (GI) cancer DNA methylation dataset that displays both serial correlations and interaction patterns. The results support and complement known aspects of DNA methylation and gene association in upper GI cancers.
READ FULL TEXT