Section: Application Domains

Change-point detection in signals

Participants : Marc Lavielle, Kevin Bleakley.

Change-point detection is historically a signal processing problem whereby we search for points at which a 1-dimensional noisy signal has an abrupt change in some way, eg. change in mean or variance. It turns out that similar methods can be developed for finding the genomic locations at which the DNA copy number changes in a cancer-stricken (or other) patient. Normally, we have two copies of DNA along the whole genome, so specific changes (gains or losses) in copy number can be associated with the specific cancer, hopefully leading to treatment possibilites. Kevin Bleakley collaborates with researchers at the Curie Institute in change-point detection in one or many DNA copy number profiles. Related to this, Marc Lavielle and Kevin Bleakley have developed methods to find change-points in histogram probability density functions using data sampled from the (unknown) density.