SillyPutty: Improved clustering by optimizing the silhouette width.

Area of Science:

• Biomedical science
• Computational biology
• Data mining

Background:

• Unsupervised clustering is crucial for analyzing complex biomedical datasets.
• Existing clustering algorithms have limitations in accuracy and speed for certain applications.

Approach:

• Developed SillyPutty, a new unsupervised clustering algorithm.
• Generated synthetic datasets using the Umpire R package for rigorous testing.
• Compared SillyPutty against established algorithms using metrics like Silhouette Width and Adjusted Rand Index.

Key Points:

• SillyPutty demonstrates comparable accuracy to state-of-the-art clustering methods as a standalone tool.
• The combination of hierarchical clustering and SillyPutty yields superior performance in both accuracy and computational efficiency.
• Performance was evaluated using multiple established metrics for robust assessment.

Conclusions:

• SillyPutty is a validated and effective method for unsupervised clustering in biomedical research.
• Hierarchical clustering followed by SillyPutty offers an optimal approach for speed and accuracy.
• This combined method provides a powerful new tool for biomedical data analysis.