Multidimensional segmentation of heterogeneous data

High-throughput methods are producing an ever increasing flood of-omics data that yield a more and more detailed and rich genomic annotation. Combining these data into coherently behaving regions lies at the heart of functional genome annotation efforts. The segmentation problem, which addresses the task of subdividing an ordered sequence of data into homogeneous, approximately constant intervals, therefore has rapidly gained practical importance in computational biology, with a strong emphasis on multi-dimensional data tracks. We suggest a new segmentation method based on decomposition thresholding, and local optimum differentiation, which detects significant breakpoints in the data to identify segment boundaries.