gms | German Medical Science

Objectives: Comparisons of paired tumor samples, i. e. primary and corresponding recurrent tumor, are performed in order to obtain insights into tumor evolution. We analyze genomic copy number aberrations measured by array comparative genomic hybridization (aCGH). A major source of bias is the difference in tumor cell content resulting in differing signal strength.

Methods: aCGH data are preprocessed by our standard analysis pipeline [1]. We use a simple parametric model of the signal intensities in the presence of an admixture of non-tumor cells [2], [1] to adjust signal intensities of the pairs to eliminate the tumor cell content bias. Adjustment parameters are estimated by robust non-linear regression. In addition we obtain quality control parameters to assess whether a pair is evaluable. After adjustment we apply our aCGH segmentation and classification algorithm [1] to both variance weighted mean and difference of adjusted data. The means essentially show the common aberrations while the differences highlight potential tumor evolution.

Results: The methods are illustrated by analyzing N = 20 pairs of primary and recurrent Glioblastoma from the German Glioma Network.

Conclusion: We present an analysis pipeline for paired aCGH samples.