Reducing system noise in copy number data using principal components of self-self hybridizations

摘要

Genomic copy number variation (CNV) is a phenomenon wherein the number of copies of a certain gene varies among individuals, causing major differences in individual characteristics and resulting in widespread genetic disorders (1). CNVs can be detected by hybridizing genomic DNA to fragments of nucleic acid (i.e., probes) used to detect target nucleotide sequences in samples (2, 3). One common method, comparative genomic hybridization (CGH), is useful but often produces irrelevant background signals, or noise. Here, we use a method known as principle component correction (principal component correction) to enhance signal in the presence of noise, improving results and reducing costs. In CGH, two genomes—an experimental sample and a reference—are hybridized to a microchip consisting of many probes. Analysis of the CGH ratio data can reveal trends in genomic regions shared by some hybridizations and not by others (Fig. P1A). These trends often interfere with CGH segmentation analysis, a common method for analyzing the genome, and lead to spurious signals. When a pair of hybridizations shares trends in one region of the genome, that pair typically has long-range correlations throughout the genome.