Methylation, deletions, and amplifications of cancer genes constitute important mechanisms in carcinogenesis. For genome-wide analysis of these changes, we propose the use of NotI clone microarrays and genomic subtraction, because NotI recognition sites are closely associated with CpG islands and genes. We show here that the CODE (Cloning Of DEleted sequences) genomic subtraction procedure can be adapted to NotI flanking sequences and to CpG islands. Because the sequence complexity of this procedure is greatly reduced, only two cycles of subtraction are required. A NotI-CODE procedure can be used to prepare NotI representations (NRs) containing 0.1-0.5% of the total DNA. The NRs contain, on average, 10-fold less repetitive sequences than the whole human genome and can be used as probes for hybridization to NotI microarrays. These microarrays, when probed with NRs, can simultaneously detect copy number changes and methylation. NotI microarrays offer a powerful tool with which to study carcinogenesis. R epresentational difference analysis (RDA) (1) and restriction fragment length polymorphism subtraction (2) were reproducibly successful in cloning deleted sequences. However, these methods are sensitive to minor impurities, are laborious, and suffer from a number of limitations (e.g., the inability to detect differences due to point mutations, small deletions, or insertions). Furthermore, the PCR amplification after the first hybridization step and before the nuclease treatment may give rise to artifacts. Excess driver DNA can result in reduced efficiency in amplification of the tester-tester duplexes because of the potential formation of residual driver-driver and drivertester duplexes that act as competitors. As RDA is based mainly on the specific amplification of the desired products and requires 95-110 PCR cycles, it suffers from a ''plateau effect'' that is characterized by a decline in the exponential rate of accumulation of amplification products. However, the major problem results from the inefficiency of the multiple restriction digestion and ligation reactions that are used in this method, and which lead to the generation of false positives. Furthermore, these experiments result in the cloning of products that usually do not represent functional genes. Similarly, the methylation-sensitive representational difference analysis (3), aimed at CpG-rich sequences, suffers from the same limitations as the original RDA.Recently, we developed a procedure for cloning deleted sequences (Cloning Of DEleted sequences, CODE) (4) that is free from some of the limitations inherent in the RDA and restriction fragment length polymorphism subtraction protocols. Our major objective was to improve the subtractive enrichment, thereby avoiding excessive PCR kinetic enrichment steps that often generate small DNA products.