Therapy-related leukemia (t-ML) is a rare but devastating complication of anticancer treatment, including treatment of acute lymphoblastic leukemia (ALL). Although a number of treatment factors have been identified that contribute to the development of t-ML, little is known about the biology of the disease or the genetic risk factors. To identify possible genetic risk factors for and characteristics of t-ML among ALL patients, we used the Affymetrix GeneChip® Human Mapping 100K set to genotype over 100,000 single nucleotide polymorphisms (SNPs) in germline DNA from 13 cases with t-ML and from 13 matched controls, as well as DNA of the t-ML blast samples of the 13 cases. Controls were matched for treatment protocol, ALL risk classification, irradiation, prior use of G-CSF, and race. Germline allele frequencies differed (at the p < 0.01 level) for 309 SNPs (203 genes) between t-ML cases and controls. Cytogenetically, most of the cases had translocations or inversions of 11q23; however, three cytogenetically-detected regions of chromosomal loss were also detected (on chromosome 5, 7, and 9, each found only once among the cases). These chromosomal deletions were all detected as regions of loss of heterozygosity (LOH) and decreased copy number by paired intra-patient germline vs t-ML blast analyses of SNP calls. However, the paired SNP data indicated 216 regions of LOH (FDR = 2.5%) corresponding to a total of 81 non-contiguous regions (Figure), most of which were not detected as losses cytogenetically, and 38 shared (among pairs) regions of LOH. As many as 45 noncontiguous and 12 contiguous LOH regions were found among those cases with no cytogenetically defined regions of loss, including patients with simple balanced translocations. In addition to cytogenetically-defined cases, using SNP analyses we detected additional t-ML cases with regions of LOH on 5q (n=1), on chromosome 7 (n=4), and on chromosome 9 (n=2). In fact, only one patient, whose cytogenetics showed a balanced translocation involving 11q23, had no extended regions of LOH detected by SNP analysis. Altogether, 5 of the 13 cases demonstrated LOH on chromosome 7. Using a change-point model, we detected 18 regions of copy number increase and 21 regions of copy number decrease in the blasts compared to paired germline DNA (p<0.01). We conclude that dense genotyping of SNPs not only accurately detects cytogenetically defined regions of gain or loss, but also chromosome abnormalities not found by conventional techniques. Genes associated with SNPs whose frequencies differed between cases and controls, with LOH, or altered copy number, are involved in apoptosis, DNA damage repair, and novel pathways, thereby providing candidate genes that may contribute to the development of t-ML. [Display omitted]