A new method to Clustering of Gene expression Data using Graph Spectrum

We can infer the function of the genes involved in the cancerous patients using clustering of genes obtained by microarray. Microarray is the device in the biologylaboratories to measure the expression levels of large numbers of genes simultaneously. In this paper, in the preprocessing stage, we extract the genes have the correlation between each other from 7129 genes we had at the first and we construct the similarity graph for the remaining graphs. The graph is the input to spectral clustering algorithm. Spectral clustering is one kind of unsupervised learning algorithm which considers the similarities among data and uses k-means in the new algebraic space. So, k-means do the clustering task in the best manner. Evaluation of clustering algorithm using Dunn index shows that this algorithm is efficient in comparison to other methods.

Keywords:

Gene Expression Data , k-means clustering , microarray , Pearson Correlation Coefficient , preprocessing , Spectrum. (key words)