Using soft consensus clustering for combining multiple clusterings of chemical structures

Abstract

The consensus clustering has shown capability to improve the robustness, novelty and stability of individual clusterings in many areas including chemoinformatics. In this paper, graph-based consensus method (cluster-based similarity partitioning algorithm CSPA) and soft consensus clustering were examined for combining multiple clusterings of chemical structures. The clustering is evaluated based on the ability to separate active from inactive molecules in each cluster. Experiments suggest that the effectiveness of soft consensus method can obtain better results than the hard consensus method (CSPA).