Clustering of Gene Expression Profiles Based on Functional Modules

Traditional clustering analysis of gene expression profiles is challenged by high measurement noise,curse of dimensionality and lacking of coherence in biological interpretations.Functional expression profiles(FEP),which is obtained by organizing the original genes expression profiles onto low-(dimension) functional modules using Gene Ontology,is proposed as new analysis indexes to cluster microarray disease samples in our novel method to tackle with the above issues.We compare the performance of hierarchical clustering and k-means clustering based on FEP and the conventional gene expression profiles(GEP) using the NCI60 dataset.The analysis results indicate that precise clustering of disease types and biological function comprehension of the analysis results can be achieved directly with FEP.In addition,FEP can also significantly reduce dimension and tackle with high with measurement noise efficiently.