Using clustering models for uncovering proteins’ structural similarity

Proteins are complex molecules that serve as building blocks in organisms, executing important tasks in order to maintain cellular environment and thus having essential roles in existence. This paper examines the usefulness of applying partitional and hierarchical clustering as unsupervised classification methods for uncovering proteins' structural similarity, based on the information contained within their conformational transitions. We investigate three representations for a protein based on the probability distributions of certain structural elements within conformational transitions and apply clustering methods to unsupervisedly classify proteins based on their structural similarity. Experiments are performed on two protein data sets and the obtained results are analyzed and compared with the results of similar existing approaches. The comparative results reveal that in many cases our proposal performs better than an earlier work in this topic.