A Simple Bayesian Approach To Detecting Changepoints Across Multiple Samples

Changepoints are abrupt changes in sequential data. The presence of multiple samples should, in theory, help to reveal subtle changepoints within noisy data. However, multi-sample changepoint detection methods are rarely used in practice because existing inference methods are complex and inefficient. In this talk, we present a simple yet effective approach to detecting changepoints across multiple samples. By transforming Bayesian multi-sample changepoint models into unconventional Hidden Markov Models, we achieve fast, closed-form approximations to the posterior distributions on changepoint indictors, segmentations, and local parameters. We present promising initial results on simulated data, and consider the problem of identifying copy number alterations in cancer biopsy samples with low tumor fractions.