Genomic instability is a widely recognized trait of cancer. Specific subchromosomal copy number changes are thought to play a driving role in the transformation of normal cells to malignant clones. These genomic copy number changes may result in deletion of one or both alleles of tumor suppressor genes, overexpression of oncogenes and rearrangements subsequently altering the transcription of target and downstream genes Several recent studies suggest that fixed genetic abnormalities in human cancers may be highly predictive of response to targeted therapeutics. This infers that the therapeutic response range in multiple cancer types could, in part, be predicted by similarities in genomic copy number profiles. High resolution array-based Comparative Genomic Hybridization (aCGH) was used to measure genome copy number in 678 cancers, including 95 breast tumors, 85 ovarian tumors, 45 head and neck tumors, 39 basal cell carcinoma tumors, 40 sarcoma tumors, 67 melanoma cell lines and 49 neuroblastoma cell lines. Additionally, aCGH data from 162 colon cancer tumors, 41 bladder cancer tumors and 55 pancreatic cancer cell lines were downloaded from publicly available databases. All data sets were normalized and regions of gain and loss were estimated for each sample. The degree of gain and loss was highly correlated for all samples (r2 = 0.89, p < 0.0001). Gain and loss frequencies were calculated for each cancer type across the genome in 1 Mb intervals. The minimum common regions of aberration for all data sets include loss of 8p, 17p, 18q and gain of 7p, 8q, and 20q. As done in previously published comparisons of cancers with expression profiles, these data were used to measure levels of similarity between cancer types. Each data set was then subject to classification using a modified hierarchical clustering algorithm that recursively searches for similarity between subsets of the genome. Globally, breast tumors, pancreatic cell lines and colon tumors appeared most similar while basal cell carcinoma tumors were consistently an outgroup. The high resolution genome copy number measurements provided by aCGH is a novel means of detecting similarity between cancer types as well as distinguishing unique common traits of specific cancer types, and is more versatile than expression profiling as it can be applied to archival specimens.

[Proc Amer Assoc Cancer Res, Volume 46, 2005]