Predicting Gene Relations Using Bayesian Networks
Aparna Sriram · OhioLink ETD Center (Ohio Library and Information Network) · 2011
Genes are the biological units responsible for the hereditary characteristics in living organisms.It takes years of arduous research to predict relations among genes using biological experiments.An alternate method that is quicker and simpler can prove to be of great significance.This research focuses on exploring a solution in the form of Bayesian Networks and its application to predict gene relations from microarray experiments.Biological pathways are used to depict the interactions among genes during biological process.In this study, such pathways were selected from the EcoCyc database for the bacterium strain EColi MG1655.For the genes involved in the pathways, the microarray data of seven experiments was obtained from the Many Microbe Microarray Database.The datasets were used for building the directed acyclic graphs using Bayesian Networks.Directed acyclic graphs (DAGs) were obtained for each experiment with a set of topological orders.A union graph was constructed based on the frequency of occurrence of each edge in the DAGs obtained.A final consensus graph was obtained by selecting a threshold frequency for the edges.A comparison of the resultant consensus graph to the biological pathway indicates that the existing gene relations can be replicated to major extent.In addition to the existing gene relations described in the pathways, a few other edges were also found to have quite a high frequency of occurrence.Previous researches based on biological experimental studies confirmed several gene relations that were revealed by these edges.The results indicate that Bayesian networks can play a vital CHAPTER I