The aim of this experiment was to define the genotype or gene expression patterns that are affiliated with colon cancer. The dataset that was analyzed was gathered from 62 patient tissue samples that were plated onto oligonucleotide arrays from which 2,000 data points were reported from each array (Alon, et al 1999).

Seven complementary genes correlated with the presence or absence of colon cancer were specific to colon epithelial cells and have been implicated in the progression of colon cancer in several studies within the literature. A surprising finding from this analysis was a gene from a Trypanosoma genus, a parasite. A literature review revealed that patients infected with this parasite demonstrate resistance to colon cancer. Patent applications have been made to protect the findings described by the genes isolated in this study, singly and in combination.

To further align our toolbox with the needs of the pharmaceutical drug development process, we, and our exclusive licensor, extended the presentation of these results using proprietary visualization tools that enable the presentation of knowledge-based decision trees for easy analysis of alternative subsets of genes of high predictive power. A binary decision model was established through iteration of the RFE-SVM over relevant subsets of promising genes obtained from unconstrained application of RFE-SVM. In this manner, potential diagnostic applications hinging on a maximum number of genes can be effectively explored. The visualization tools allow quick identification of relevant findings through easily perceived node color labeling and edge magnitude in the resulting diagrams. Additional constraints, such as gene patentability, ease of measuring a protein in serum, estimation of value as a drug target, or economic viability could be introduced to modify the preferred path in real time.