Comprehensive characterization of genomic alterations in cancer cell lines will advance our understanding of cancer biology, and could also provide a basis for choosing relevant cell line models to study a particular aspect of cancer disease biology,LMT-28 or to screen for antagonists of certain cancer pathways. To evaluate NGS technologies and to characterize genomic mutations in cancer cell lines, we have analyzed data from the Roche Nimblegen exome capturing array and Roche 454 NGS technologies, applied to eight commonly used cell lines representing several major cancer types. We demonstrate that exome sequencing can be a reliable and cost effective way for identifying genomic alterations in cancer genome, and generated a comprehensive catalogue of genomic alterations in coding regions of eight cancer cell lines. We detected on average 14,340 sequence variants per cell line. The majority of these differences are known polymorphisms in normal human population. On average 2,779 variants per cell line are not found in the dbSNP database, and therefore represent novel sequence variations and/or somatic mutations.On average 1,904 of the 2,779 novel variants are non-synonymous, i.e. they alter codon specificity. These variants are more likely to change protein functions and impact cellular phenotypes. Deletions or amplifications of chromosomal segments are common alterations in cancer genomes. In principle, the sequencing read depth in a region should be proportional to its copy number. However,Brassinin the relatively modest read depth of the current study could give undue weight to random variations in read depth. Variability in read depth could also arise from technical aspects of the exome sequencing process. For example, the exome capturing array could vary in efficiencies for different exon regions due to diverse sequence composition. To assess the possibility of estimating copy number information from our exome sequencing data, we compared average sequence read depths with copy-number data estimated from SNP6 platform. As show in Figure 2, there is a positive correlation between sequence read depth and copy-number, with Pearson correlation coefficient of 0.41. The variation in read depth makes it challenging to accurately detect low-level copy-number changes.