Integrating gene expression, protein interaction and protein domain data to improve gene expression clustering

Abstract

ABSTRACT Cell activity is carried out by the interaction of various proteins. Complex interactions among proteins constitute molecular pathways, which are the mechanisms by which the living cells perform biological processes. Understanding pathways is crucial in revealing mechanisms of cellular activity, thus understanding the reasons behind genetic disorders. Domains, which are independent subunits of proteins, play an important role in protein interactions. The first method presented in this thesis uses association rule mining on protein interaction data to extract domain-domain interaction rules. The method was applied on a database of protein interactions, which resulted in rules, some of which are supported by biological knowledge. Microarray expression data is another data source to study protein interactions. Most microarray data analysis methods are based on clustering genes that show similar expression patterns. However, clustering results often need to be refined, which can be done either by using biological expertise or by integrating other biological data. The second proposed method integrates domain-domain interaction rules with microarray data. The method is based on a previously developed probabilistic model which unifies protein interaction and microarray data. Results show that integrating domain-domain interaction rules produces gene clusters of higher coherence. Finally, a paxallelization of the second proposed method and its implementation, together with performance results are presented. m