A novel Gene Clustering Algorithm based on the Integration of Expression Data and Functional Profiles

The study of gene clustering algorithms from available experimental data is one of most challenging tasks in system biology. It’s the basic study when researching genes and their relative relations. Traditional gene clustering algorithms are based on gene expression data only. But for most cases, these algorithms can’t find out all possible genes relationships. A novel gene clustering algorithm which integrates biological messages, such as functional profiles, into gene expression data was proposed in this paper. A novel distance measurement amongst genes mixed gene expression data and biological messages together was proposed in our work for more accuracy calculation. K-means as a traditional clustering algorithm was used to get clustering results. Three optimal algorithms which contained GA, PSO and the novel optimal algorithm GFA were used in K-means algorithm. The proposed algorithm is validated on both the simulated genes data and real benchmark genes data in gene database. And the results were used to compare with each other. The cross-validation results confirmed the effectiveness of our algorithm, which outperforms significantly other previous algorithms.