Analysis of arrayCGH Data for Estimation of Genetic Tumor Progression - Ph.D. application talk

a meaningful tumor classification and selecting appropriate therapies. The
accumulation of genetic alterations during tumor progression can be used for
the assessment of the genetic status of the tumor. For modelling dependencies
between the genetic events, evolutionary tree models have been used.
ArrayCGH data is now available for various types of cancer. Our goal is to
automatically identify specific chromosomal alterations in brain tumor and
prostate cancer arrayCGH data, and then compute the corresponding oncogenetic
trees. Many good algorithms have been proposed for estimating the
copy number of DNA regions from noisy arrayCGH data. We chose GLAD,
which uses an adaptive weights smoothing algorithm to detect breakpoints of
segments with constant copy number. Once the copy number is estimated,
we introduce a method to statistically decide whether a region is gained,
normal or lost. In order to determine the genetic profile of each tumor sample,
we propose an algorithm to detect the mutations that are common to
a large proportion of them, thus providing with all the data necessary for
computing the genetic trees. Our results are consistent with what was previously
known about the two types of cancer, but they bring a higher precision
due to the high resolution of arrayCGH data. The genetic profiles can be
subject to further interesting analysis. Clustering of tumor samples of different
types before and after the arrayCGH filtering shows that our method
extracts meaningful information.