A comparative study of clustering algorithms for protein sequences

With the development of sequencing technologies, more and more protein sequences are uncharacterized. Clustering protein sequences into homologous groups can help to annotate uncharacterized protein sequences. In recent years, many clustering algorithms have been proposed to analyze protein sequences. It may be necessary to perform a comparative study of these algorithms, and help biologists to choose suitable clustering algorithm for their tasks. In this work, we present a comparative experiment on three clustering algorithms: BlastClust, Spectral clustering, and TribeMCL. We conducted two types of experiment for each algorithm :(1) Default parameters experiment; (2) Parameters tuning. The results of evaluation uncover that TribeMCL outperform the other methods. BlastClust is extremely dependent on the selection of parameters values.