Analysis of High-throughput DNA Methylation Bead Arrays Utilizing Bayesian Genotyping Algorithms

We present a statistical framework, MAMS-M, for determining the methylation status of hundreds of cancer related CpG sites. MAMS-M extends and adapts our previous SNP genotyping algorithm, MAMS, to methylation bead array data, exploiting the similarities in data structure between the two platforms. MAMS-M employs a multi-site, multi-array model-based clustering approach to derive initial methylation calls, and then recalibrate these calls and associated confidence measures using site-specific adjustments. We demonstrate the performance of MAMS-M using a real-life data set with cancer applications.