| My research focused on the classification and functional genomics of mouse hepatocellular carcinoma (HCC). I use three approaches to evaluate and address three hypotheses. First, I demonstrated that, with high accuracy, I could classify tumors according to their initiating oncogene(s). I also determined that, of three machine learning algorithms (Support Vector Machine or SVM, Naive Bayes, and Random Forest), SVM performed the best with my data. This study showed that liver tumors display distinct gene expression patterns that can identify their initiating oncogene(s).;Second, I used clustering analysis to examine the extent of similarities within a genotype (oncogene-specific) group and between genotypes. Clustering of the liver tumors lead to functional evaluation, in which I correlated functional categories of changed gene expression with phenotypic data on tumor aggressiveness. I identified several pathways that correlated with either slow or fast growing c-Myc combinations, and some that might explain the aggressiveness of H-ras combined with beta-catenin. In addition, I also identified candidate genes that correlated with the gender effects observed in the phenotype.;Third, gene expression was compared in normal liver, liver in which oncogene(s) were expressed for up to two weeks, and liver tumors. I found that 10 fold more gene expression changes were acquired during the tumorigenesis process then were present after short term expression of the initial molecular alteration(s). Examining the sequence of expression changes allowed me to propose a link between certain expression patterns and specific categories of gene function. These new insights improve our understanding of the molecular mechanisms of HCC. |
