Identification of significant regional genetic variations using continuous CNV values in aCGH data

Abstract

Array comparative genomic hybridization (aCGH) provides a technique to survey the human genome for chromosomal aberrations in disease. The identification of genomic regions with aberrations may clarify the initiation and progression of cancer, improve diagnostic and prognostic accuracy, and guide therapy. The analysis of variance (ANOVA) model is widely used to detect differentially expressed genes after accounting for common sources of variation in microarray analysis. In this study, we propose a method, shifted ANOVA, to detect significantly altered regions. This method, based on the standard ANOVA, analyzes changes in copy number variation for regions. The selected regions have the group effect only, but no effect within samples and no interactive effects. The performance of the proposed method is evaluated from the homogeneity and classification accuracies of the selected regions. Shifted ANOVA may identify new candidate genes neighboring known because it detects significantly altered chromosomal regions, rather than independent probes.