Array preparation labeling hybridization and data analysis

Many of the basic procedures followed in microarray-based genome profiling are similar, if not identical, to those followed in expression profiling, including the use of specialized microarray equipment and data-analysis tools. Since microarray-based expression profiling has been well established over the last decade, much can be learned from the technical advances made in this area. However, there are also distinct differences such as target and probe complexity, stability of DNA over RNA, the presence of repetitive DNA and the need to identify single copy number alterations in genome profiling.

Specifically, the array CGH procedure includes the following steps. First, large-insert clones such as BACs are obtained from a supplier of clone libraries such as BACPAC Resources at the Children's Hospital Oakland Research Institute ( Then, small amounts of clone DNA are amplified by either degenerate oligonucleotide-primed (DOP) PCR (6) or ligation-mediated PCR (7) in order to obtain sufficient quantities needed for spotting (~1 ^g/^l). Next, these PCR products are spotted onto glass slides coated with substrates such as aminosilane using microarray robots equipped with high-precision printing pins. Depending on the amount of clones to be spotted and the space available on the microarray slide, clones can either be spotted once per array or in replicate. Repeated spotting of the same clone on an array increases the precision of the measurements if the spot intensities are averaged, and allows for a detailed statistical analysis of the quality of the experiments. Patient and control DNAs (100 ng-1 ^g) are usually labeled with either Cy3 or Cy5-dUTP using random priming and are subsequently hybridized onto the microarray in a solution containing an excess of Cotl-DNA to block repetitive sequences. Hybridizations can either be performed manually under a coverslip, in a gasket with gentle rocking or, automatically using commercially available hybridization stations. These automated hybridization stations allow for an active hybridization process, thereby improving the reproducibility as well as reducing the actual hybridization time, which increases throughput. The hybridized DNAs are detected through the two different fluorochromes using standard microarray scanning equipment with either a scanning confocal laser or a charge coupled device (CCD) camera-based reader, followed by spot identification using commercially or freely available software packages.

The increase in data obtained through high-density arrays requires standardized storage systems as well as thorough statistical tools, similar to those required for microarray-based gene expression profiling (8, 9). Owing to the complicated process of producing and hybridizing spotted micro-arrays, a certain degree of systematic variation does exist in the data produced. Normalization of microarray data is used to eliminate such systematic variation and, therefore, represents an important preprocessing step in the analysis of almost all microarray data. One of the most frequently applied normalization procedures in microarray-based expression studies is the locally weighted scatterplot smoothing (LOWESS). Several laboratories have successfully introduced this procedure in array

CGH applications (10-12). After data normalization, automated statistical procedures are required for reliable detection of genomic copy number changes. One such algorithm is the Hidden Markov Model (HMM) which, in our hands, is not only suited for distinguishing genuine copy number changes from random microarray noise, but also for precisely localizing the start- and end-points of each copy number alteration (unpublished results; see Figure 11.1). Finally, digitized intensity differences in the hybridization patterns of the DNAs onto the cloned fragments can be interpreted as copy number differences between the test and reference genomes.

This technique, once established and validated, allows high-throughput DNA copy number screening with a resolution limited only by the size of the clone fragments used (currently ~100 kb using BAC arrays).

0 0

Post a comment