Large-scale profiling of the Arabidopsis transcriptome.

DNA microarray is a powerful technology for parallel analysis of gene expression (Brown and Bostein, 1999). Since microarray technology emerged 5 years ago (Schena et al., 1995), the number of genes that can be monitored by this technology has increased from several hundreds (Yuan et al., 1998; Aharoni et al., 2000; Reymond et al., 2000) to several thousands (Arabidopsis Functional Genomics Consortium Microarray, http://afgc.stanford.edu/afgc html). At Novartis Agricultural Discovery Institute, Inc. (NADII), microarray is an important component in our toolbox for transcription profiling. In addition, we have developed and are using other gene expression monitoring technology platforms for gene expression profiling (emphasizing coverage) and gene expression diagnostics (emphasizing throughput). These technologies include serial analysis of gene expression, cDNA fingerprinting, and microbead-based liquid microarray. Here we focus on the microarray technologies used at NADII. It should be noted that there are at least two nomenclature systems that have been used to describe the hybridization partners in the microarray field. Consistent with the common nomenclature (Duggan et al., 1999; Lipshultz et al., 1999; Southern et al., 1999), we use “probe” to refer to the tethered nucleic acid molecules used to interrogate the experimental samples and “target” to refer to the free hybridization partner in the experimental sample. Our gene expression microarray program has two technology platforms: oligonucleotide-based probe array (GeneChip) and cDNA-based array. For the purpose of gene discovery using Arabidopsis as a model system, a large-scale profile of its transcriptome is needed. We selected the oligonucleotidebased array (Lockhart et al., 1996) as our primary platform technology because of the following reasons. First, the transcript abundance of each gene can be accurately measured by multiple probe pairs. Second, the data can be produced with a moderate throughput and a large scale. The designed arrays are commercially manufactured using photolithography technology, therefore, the time-consuming and labor-intensive array fabrication process is eliminated. Moreover, human errors that often occur during the clone tracking process are also eliminated. Third, standardized data produced by the array can be easily normalized and interrogated. This is important when cross-project comparison is needed. And finally, the necessary genomic information for oligonucleotide probe selection is available from the Arabidopsis genome sequencing project (Lin et al., 1999; Mayer et al., 1999). To profile the Arabidopsis transcriptome on a large scale, in addition to designing a high-density oligonucleotide probe array, we also tested and developed protocols for sample preparation; we developed the laboratory information management system (LIMS) for project, sample information, and data management; and we developed and integrated a number of analysis tools for data mining.