A universal feature of cancer cells is genomic instability, which is thought to be required. Following this paradigm, it is now understood that genomic instability can arise from defects in DNA synthesis and repair, chromosome segregation, checkpoints, telomere loss and other biological processes that result in point mutations, copy number variation and gain/loss of biological functions. Hepatocellular carcinoma is the second most prevalent cancer of Asian populations and the third leading cause of cancer death in the world. Currently the only effective treatment option is surgery. HCC commonly arises in patients with viral hepatitis and/or cirrhosis where extensive inflammation exposes hepatocytes to mitogenic stimuli. The pre-neoplastic phase is characterized by a number of changes, including the emergence of telomere shortening and the appearance of genomic alterations. Structural changes in the genome progressively accumulate during the transition to neoplasia and from early to late stage HCC. Genomic alterations in HCC are heterogeneous in that many loci have been reported to be altered but generally at a low prevalence. This leads to the hypothesis that there are alternate perturbations that promote tumorigenesis in HCC. Integrative genomics analysis has been successfully applied to many non-cancer diseases and has described networks of gene variation by testing all possible associations across diverse populations segregating the disease of interest. This work has established that genes are generally part of coherent networks, and that the most significant associations of genes to disease often occur in the context of network sub-regions where many or all members of these sub-networks are associated with each other and with disease traits. Such sub-networks have further been associated with DNA variation and 3,4,5-Trimethoxyphenylacetic acid validated as causally driving disease outcome. Here we have examined gene network structure using a collection of,250 matched tumor and adjacent normal samples removed from HCC patients during surgical resection and have assessed whether these networks are associated with DNA and disease variation in the HCC cohort. The approach was in essence to uncover interactions within and between the data types measured in this population in AN and TU tissues in an open ended, comprehensive and completely data driven manner. The interactions characteristic of tumors were compared to normal tissue to reveal tumor specific changes. Here we present the results of that comprehensive analysis and show that sCNV robustly alters the expression of a large number of genes and also the relationship of those genes to survival in either AN or TU tissue, and that tumorigenesis largely involves disruption of normal functions and the activation of a smaller set of functions that may be critical to disease progression. The data suggested that genes predictive of survival in AN tissue may be rate limiting steps for tumorigenesis. Consistent with this hypothesis a treatment that induces HCC tumorigenesis in mice, MET oncogene overexpression, was found to selectively alter the expression of genes predictive of survival in AN tissue of humans. To assess whether the differential correlations were randomly distributed amongst the significant 4-(Benzyloxy)phenol gene-gene correlations or whether there was some higher level structure, we examined the distribution of the number of differential correlations for each gene. We observed that whereas most genes participated in a small number of differential correlations, there was a subset of genes that participated in many differential correlations.