The Genom of Homo sapiens.pdf

324 LI ET AL.Figure 1. Array-based CGH to detect copynumber variations. Cloned DNA samples representingdifferent genomic regions are printed onglass surfaces to make microarrays. Test andnormal control genomic DNA is differentiallylabeled with dye-labeled nucleotides and co-hybridizedto the arrays. The copy number variationsalong a chromosome are represented bythe fluorescence ratios of test DNA to controlDNA different from 1 (usually 1.2 is consideredas a cutoff for gain, and 0.8 as a cutoff for loss).Large clones, mostly available as BACs, are preferred formaking CGH arrays because they give good sensitivity,and information on mapped clones is easily accessible inthe human genome databases (http://genome.ucsc.edu/;http://genome.wustl.edu/projects/human/). cDNA arrayshave been demonstrated to be useful for CGH (Pollack etal. 1999; Heiskanen et al. 2000; Beheshti et al. 2003) buthave several limitations which make them less attractivefor detection of DNA copy number changes. First, theydo not generate sufficient sensitivity for complex genomicprobes because of their small size (0.5–3 kb).cDNA clones normally generate hybridization signals30–80 times weaker than BAC clones for human ormouse genomic probes. This lack of sensitivity requiresan additional signal amplification step (Heiskanen et al.2000). Second, large genomic regions sparsely coveredby cDNA clones will give poor resolution, and those genomicregions lacking genes will not be probed at all. Ouranalysis of the complete genomic sequence indicates thata total of 620-Mb euchromatic genomic sequences belongin regions larger than 500 kb without any annotatedgene. These gene-barren regions may contain long-rangegenomic regulatory elements such as locus control regions,the deletion of which will affect multiple genes.Third, for applications that require high-level amplificationof genomic DNA probes, amplification bias willmore likely cause inaccurate copy number detection withcDNA arrays than BAC arrays because the larger size ofBACs is more effective in averaging out amplificationbias.One apparent disadvantage of using BAC clones forCGH-based detection of DNA copy number changes isthat some BACs contain significant amounts of repetitivesequences. Highly repetitive sequences in BACs can usuallybe effectively suppressed by Cot I DNA blocking.The most problematic BAC clones, however, are thosecontaining high levels of low-copy repeats or similarcross-hybridizing sequences. These clones, in theory,will show false-positive or false-negative gains andlosses, respectively (Fig. 2). If, for example, a genomicregion is gained in the patient, false-positive gains arealso caused by cross-hybridization of DNA from this regionto other homologous regions in the genome. On theother hand, cross-hybridization may generate false-negativebut not false-positive loss detection. False-negativeloss detection is a major limitation that reduces the realresolution of BAC arrays, making some small genomicregions noninformative. A complete solution for theFigure 2. Models of false-positive and -negativedetection caused by cross-hybridization. (A)Cross-hybridization causing false-positive gains.Gain of sequences in region a in the test samplecan spill to cross-hybridizing regions b and c,causing false-positive gain detection in b and c.(B) Although real loss of sequences in region aexists, it is not detected due to partial or completecross-hybridization of sequences from regions band c, causing false-negative loss detection in regiona.