Clustering gene-expression data with repeated measurements.

Area of Science:

• Bioinformatics
• Computational Biology
• Statistical Genomics

Background:

• Clustering is a fundamental technique for analyzing high-dimensional array data.
• Repeated measurements are increasingly common in biological array datasets.
• Existing clustering methods may not fully leverage the information from repeated measures.

Purpose of the Study:

• To evaluate clustering algorithms for array data with repeated measurements.
• To identify algorithms that enhance cluster accuracy and stability.
• To demonstrate the benefits of incorporating repeated measures into clustering.

Main Methods:

• Comparative analysis of various clustering algorithms.
• Implementation of algorithms designed to handle repeated measures.
• Evaluation using metrics for cluster accuracy and stability.
• Focus on an infinite mixture model with an integrated error model.

Main Results:

• Algorithms exploiting repeated measurements significantly outperform those that do not.
• Enhanced accuracy and stability were observed in clusters generated by methods using repeated measures.
• The infinite mixture model-based approach demonstrated superior performance.

Conclusions:

• Incorporating repeated measurements into clustering algorithms is crucial for accurate array data analysis.
• The infinite mixture model with an error model provides a robust and effective approach.
• This methodology enhances the reliability of biological data interpretation.