A statistical approach for CGH microarray data analysis
Abstract
Microarray-CGH experiments aim at detecting and mapping chromosomal imbalances, by hybridizing targets of genomic DNA from a test and a reference sample. Probes are constituted by sequences of genomic DNA (BACs) that are mapped on the genome. For this reason, the signal has a spatial coherence that has to be handled by specific statistical tools. Process segmentation seems to be a natural framework for this purpose. A CGH profile can be viewed as a succession of segments that represent homogeneous regions in the genome whose BACs share the same relative copy number in average. We model a CGH profile by a random gaussian process whose distribution parameters are affected by abrupt changes at unknown coordinates. Two major problem arise: the estimation of the break-points coordinates and the estimation of the number of segments. A dynamic programming algorithm is used to partition the data into a finite number of segments. A model selection approach is used to determine the number of segments in the profile, using an adaptative method. We explain why classical penalized criteria can not be used in the context of break-points detection and show the potentialities of our methodology, using publicly available data sets. We detect previously mapped chromosomal aberrations and disuss the performance of our methodology on noisier data concerning breast cancer cell lines.