This indicates that certain genes represent hub nodes in the differentially connected matrix that arose from tumorigenesis and as such may be of particular importance. Given the large scale changes in expression and correlation structures arose during the process of tumorigenesis, we sought to identify the causal drivers of these changes. Somatic copy number variation is a common feature of many solid tumor types and has been associated with the aggressiveness of disease. For HCC in particular sCNV has been observed at the earliest stages of disease and increases in prevalence with disease progression. We therefore assessed the prevalence of sCNV in HCC and to what extent it was associated with gene variation in the TU tissue. DNA variation was assessed in the AN and TU samples using Illumina high-density SNP microarrays. sCNV were estimated using smoothed logR ratio’s of adjacent markers at 32,711 evenly spaced loci through the genome. In the TU samples evidence of frequent amplification or deletion involving large genomic regions was seen. In contrast very few such events were observed in the AN samples with this analysis. sCNV variation was compared to gene variation in both the AN and TU samples. Consistent with previous studies of other cancer types and radiation hybrids, Atropine sulfate strong positive correlations between genes and sCNV markers were identified in cases where the corresponding genes overlapped or were near the sCNV marker being tested, referred to here as cis-acting associations. The most likely explanation for this observation in TU tissue is that sCNV induce proportional changes in genes that were proximal to the site of that sCNV. In contrast there were no cis-acting associations between AN CNV markers and AN genes beyond what would be expected by chance, indicating that the ciscorrelations between sCNV and expression were tumor specific. Given this correlations to copy number variation were only investigated using TU tissue. Consistent with this it has recently been reported that the structure of sCNV is frequently shared across multiple tumor types. This might suggest that the cis and trans correlations reported here in HCC and cells in culture may be relevant to many tumors types with shared sCNV structure. The absence of evidence indicating direct communication between AN and TU tissues leaves the possibility that the AN tissue retains to a significant degree the characteristics of the pretumor cells from which the tumor evolved. In this case, the variance of genes prior to tumorigenesis predicted the probability of tumorigenesis occurring, but after that process had occurred the same genes were no longer predictive. Consistent with this, once the AN network was transformed by changes in gene-gene correlations driven by sCNV, the formerly predictive genes would no longer be predictive. Since the process of tumorigenesis is linked and perhaps driven by network transformation, genes predictive of that process were also predictive of survival. We can derive from this hypothesis a testable prediction. If the starting state of the gene networks is a determinant of the likelihood of tumorigenesis occurring, then treatments that promote tumorigenesis should selectively alter genes that participate in the network transformation that characterizes that process. To this end we took advantage of a genetic model of HCC where the oncogene MET was over-expressed in the Catharanthine sulfate livers of mice, resulting in a large increase in the numbers of HCC tumors for that strain. The hypothesis above predicts that a treatment that promotes HCC tumorigenesis, should, prior to the onset of tumorigenesis.