Analysis of microarray data with multiple phenotypes

Microarrays are widely used to monitor gene expressions to yield information for genomes and to inquire relations between gene expression levels and multiple phenotypes. The relation between the genotype and the phenotype is also not one-directional. Rather, genotypes and phenotypes are interacting and mutually dependent. The complicated relations call for new statistical solutions to handle the multi-phenotype data. The first objective of this research is to use Factor Analysis to discover the independent factors behind the multiple phenotypes. The factors can be then used to test the differentially expressed genes separately. The second objective of this research is to develop a genotype-phenotype interaction model to test the differentially expressed gene. With thousands of genes and tens of phenotype, the number of available microarrays is relatively small and causes breakdown of conventional multivariate analysis methods. We´ll also attempt to resolve this small-sample-size problem often seen in microarray data analysis. Hepatoma and Breast cancer datasets will be used to demonstrate and validate the proposed methods.