Performance evaluation of a community structure finding algorithm using modularity and C-rand measures

Abstract

Biological networks, social networks, and the World Wide Web are some examples of real world networks exhibiting community structure. We present a concise review of community structure finding (CSF) algorithms and applications. We apply a CSF algorithm and various other algorithms on three different microarray data sets. We calculate modularity and C-rand indices as an indication of the quality of each clustering of the three data sets. We compare the performance of the CSF algorithm with the performance of three other algorithms: hierarchical clustering (HC) algorithm, K-means, dynamic tree cut (DTC) algorithm and Naive Bayes Clustering (NBC) using both C-rand and modularity values.We report that the CSF algorithm detects clusters resulting in high modularity; however the CSF does not result in clusters with high C-rand values compared to the other methods.

Performance of an Ensemble Clustering Algorithm on Biological Data Sets

“…The value of average hits per gene and average hit percentage in cluster is 12.5 and 9.05 respectively. These numbers appear to be rather high considering the analysis that we performed in [5], where we carried out a similar analysis for clusters of genes of Breast B data set created by CSF and DTC algorithms. Since we have not performed Chilibot analysis for all the clusters in this paper, we cannot make a fair comparison of our available results with the results in [5].…”

“…These numbers appear to be rather high considering the analysis that we performed in [5], where we carried out a similar analysis for clusters of genes of Breast B data set created by CSF and DTC algorithms. Since we have not performed Chilibot analysis for all the clusters in this paper, we cannot make a fair comparison of our available results with the results in [5]. However, at least with respect to the result corresponding to cluster 1, it seems that there are a considerable number of associations between the genes in this cluster based upon the PubMed database.…”

“…We wanted to see if the ensemble clusters obtained using the ensemble method generally agreed with the biological literature using the "Search for relationships between many genes, proteins, or keywords" option provided by Chilibot [25], [26], which is a freely available online tool that searches abstracts from the PubMed literature database [27] for specific relationships between proteins, genes, and keywords [5]. In this preliminary analysis, we examined only one cluster obtained from the Breast B data set.…”

“…We searched for the keyword "Cancer" along with the gene names contained in cluster 1 as we did in [5]. As output, Chilibot provided the number of PubMed literature database abstracts containing the keyword "Cancer" associated with a particular gene, along with the number of gene interactions, i.e., the number of abstracts where two genes appear together, for each gene.…”

“…The base clustering algorithms used for the ensemble approach used here are hierarchical clustering (HC), K-means, dynamic tree cut (DTC), fuzzy C-means and a community structure finding algorithm (CSF). All of these algorithms except for fuzzy Cmeans were used and detailed in a previous study [5] to compare the performance of the individual algorithms with one another.…”