Initial points selection for clustering gene expression data: a spatial contiguity analysis-based approach.

Area of Science:

• Bioinformatics
• Computational Biology
• Data Science

Background:

• Clustering is vital for gene expression data analysis.
• Iterative clustering methods, such as k-means, are sensitive to initial conditions, leading to local minima and inefficiency.
• Current methods struggle with accurate initial point selection.

Purpose of the Study:

• To propose a novel spatial contiguity analysis-based approach for clustering gene expression data.
• To address the sensitivity of iterative clustering techniques to initial starting conditions.
• To enhance the accuracy and computational efficiency of gene expression data clustering.

Main Methods:

• Utilizing spatial contiguity analysis to guide initial point selection.
• Employing principal component analysis (PCA) to identify representative initial points from distinct clusters.
• Applying the proposed method to benchmark datasets for validation.

Main Results:

• The proposed approach effectively identifies suitable initial points for clustering.
• Demonstrated avoidance of local minima issues common in iterative clustering.
• Showcased accelerated computation times compared to standard methods.
• Validated effectiveness across multiple benchmark datasets.

Conclusions:

• The spatial contiguity analysis-based approach offers a robust solution for gene expression data clustering.
• PCA-enhanced initial point selection significantly improves clustering performance and efficiency.
• This method provides a valuable advancement for analyzing complex biological datasets.