Recurring chromosome abnormalities are strongly associated with certain subtypes of leukemia, lymphoma and sarcomas. More recently, their potential involvement in carcinomas, i.e. prostate cancer, has been recognized. They are among the most important factors in determining disease prognosis, and in many cases, identification of these chromosome abnormalities is crucial in selecting appropriate treatment protocols. Chromosome translocations are frequently observed in both de novo and therapy-related acute myeloid leukemia (AML) and myelodysplastic syndromes (MDS). The mechanisms that result in such chromosome translocations in leukemia and other cancers are largely unknown. Genomic breakpoints in all the common chromosome translocations in leukemia, including t(4;11), t(9;11), t(8;21), inv(16), t(15;17), t(12;21), t(1;19) and t(9;22), have been cloned. Genomic breakpoints tend to cluster in certain intronic regions of the relevant genes including MLL, AF4, AF9, AML1, ETO, CBFB, MYHI1, PML, RARA, TEL, E2A, PBX1, BCR and ABL. However, whereas the genomic breakpoints in MLL tend to cluster in the 5' portion of the 8.3 kb breakpoint cluster region (BCR) in de novo and adult patients and in the 3' portion in infant leukemia patients and t-AML patients, those in both the AML1 and ETO genes occur in the same clustered regions in both de novo and t-AML patients. These differences may reflect differences in the mechanisms involved in the formation of the translocations. Specific chromatin structural elements, such as in vivo topoisomerase II (topo II) cleavage sites, DNase I hypersensitive sites and scaffold attachment regions (SARs) have been mapped in the breakpoint regions of the relevant genes. Strong in vivo topo II cleavage sites and DNase I hypersensitive sites often co-localize with each other and also with many of the BCRs in most of these genes, whereas SARs are associated with BCRs in MLL, AF4, AF9, AML1, ETO and ABL, but not in the BCR gene. In addition, the BCRs in MLL, AML1 and ETO have the lowest free energy level for unwinding double strand DNA. Virtually all chromosome translocations in leukemia that have been analyzed to date show no consistent homologous sequences at the breakpoints, whereas a strong non-homologous end joining (NHEJ) repair signature exists at all of these chromosome translocation breakpoint junctions; this includes small deletions and duplications in each breakpoint, and micro-homologies and non-template insertions at genomic junctions of each chromosome translocation. Surprisingly, the size of these deletions and duplications in the same translocation is much larger in de novo leukemia than in therapy-related leukemia. We propose a non-homologous chromosome recombination model as one of the mechanisms that results in chromosome translocations in leukemia. The topo II cleavage sites at open chromatin regions (DNase I hypersensitive sites), SARs or the regions with low energy level are vulnerable to certain genotoxic or other agents and become the initial breakage sites, which are followed by an excision end joining repair process.