Systematic scan and statistical analysis of loss of heterozygosity (LOH) has been widely used to define chromosomal aberrations in various cancers for cloning of tumor suppressor genes and for development of prognostic markers. However, the establishment of novel strategies is needed, so that the nonrandom but heterogeneous chromosomal aberration data could provide significant insights into our understanding of molecular pathogenesis of cancers. After comprehensive allelotyping of recurrent allelic losses with 441 highly informative microsatellite markers and overlapping LOH regions on human hepatocellular carcinoma (HCC) chromosomes, 33 minimal deleted regions (MDRs) were revealed. Five and 15 of the 33 MDRs have physical intervals in less than 5 and 10 Mb, respectively, with the smallest MDR9p1 of 2.2 Mb located at 9p21.3-p21.2. Statistical and Kaplan-Meier survival analysis revealed a significant association between the loss of MDR15q1 (15q21.1-q22.2) and the HCC patient survival (adjusted P = 0.033). After cluster analysis of 33 MDRs that represented LOH profiles of each HCC tissue based on clinicopathological features and p53 mutations, two major genetic pathways, low-stage and advanced-stage HCC, were uncovered based on high concordance of MDR clusters. We propose that the definition of genome-wide MDRs on the cancer genome not only narrows down the location of existing tumor suppressor genes to facilitate positional candidate cloning and develop potential prognostic markers after statistical association of MDRs with clinicopathological features but also dissects genetic interactions and pathways of chromosomal aberrations in tumorigenesis.