Searching for Significant Genes in Cancer Metastasis by Tissue Comparisons

摘要

DNA Microarrays allow scientists to simultaneously measure the expression levels of thousands of genes. However, an important need arises as to identify those genes closely associated with a particular state of interest, such as cancer, in order to discover useful biological information and efficiently classify new samples. The identification of marker genes is often based on the differential expression of groups of genes and/or their predictive potential manifested in classification experiments. Important aspects that need to be verified on both biological and statistical grounds are the actual problem considered and the algorithmic method selected. In this paper we consider the question of gene differentiation in cancer and how it can be explored through blood sample analysis. We study two algorithmic approaches based on support and relevance vector machines. The results indicate that the latter concept performs better in the specific biological environment, extracting meaningful biological concepts.