On the analysis of microarray data with applications to cancer and cardiovascular disease

A microarray experiment exploits the association between global gene expression and phenotype to identify genes and pathways that enable a faithful and parsimonious modeling of the phenotypic variation observed in complex biological systems. Statistical methods are used to identify significant genes based on the degree of correlation between the gene's expression profile and observed changes in phenotype as measured on the replicate samples of the microarray experiment. The resulting list of significant genes is then reduced to pathways and functional clusters that are statistically over-represented in on the list in order to isolate those significant genes that are most relevant to the observed variation in phenotype. The internal structure of the most significant pathways and functional gene clusters provide hypotheses regarding gene interactions that provide mechanisms that explain the observed changes in biological state. Moreover, the expression profiles of significant genes provide features (i.e., explanatory variables) that are able to directly model and generalize the phenotypic dynamics observed in the experiment to a wider sample population.