Crimp: An efficient tool for summarizing multiple clusterings in population structure analysis and beyond.

Area of Science:

• Computational Biology
• Population Genetics
• Data Analysis

Background:

• Unsupervised clustering techniques can produce clusterings with different cluster orders, a phenomenon known as label switching.
• Label switching complicates the comparison and summarization of replicate analyses, particularly in population structure studies using multilocus genotype data.
• Existing postprocessing tools for label switching have limitations, especially with large datasets.

Purpose of the Study:

• To introduce Crimp, a novel command-line tool designed to address the label switching problem in clustering.
• To provide a fast, scalable, and effective solution for aligning clusters across replicate analyses.
• To facilitate accurate comparison and summarization of clustering results, especially for large population genetics datasets.

Main Methods:

• Crimp employs a heuristic algorithm for aligning clusters across replicate clusterings with the same number of clusters.
• An exact algorithm is available for smaller problem sizes.
• The tool supports row-specific weights and offers input/output compatibility with CLUMPP, including objective function evaluation.

Main Results:

• Benchmark analyses demonstrate that Crimp outperforms alternative tools in terms of runtime and quality measures, particularly for large datasets.
• Crimp effectively mitigates label switching, enabling more meaningful comparisons of population structure.
• The tool's computational efficiency allows application to large datasets while maintaining competitive solution quality.

Conclusions:

• Crimp is a valuable and efficient tool for correcting label switching in replicate clusterings.
• Its scalability and performance make it suitable for large-scale population genetics analyses.
• Crimp facilitates improved comparison and averaging of clustering results, enhancing the interpretation of population structure.