Clustering of change patterns using Fourier coefficients.

Area of Science:

• Bioinformatics
• Computational Biology
• Systems Biology

Background:

• Understanding gene expression dynamics is crucial for identifying biologically related gene groups.
• Clustering gene expression data based on temporal change patterns presents unique challenges due to functional complexity.
• Derivative Fourier coefficients offer a dimension-reducing approach capturing underlying function and statistical properties.

Purpose of the Study:

• To develop a statistical model for clustering gene expression data based on similar temporal change patterns.
• To leverage derivative Fourier coefficients for robust gene expression pattern analysis.
• To identify gene groups with shared biological properties through pattern clustering.

Main Methods:

• Developed a model-based clustering approach utilizing derivative Fourier coefficients.
• Estimated gene expression change patterns using Fourier series representation.
• Applied a multivariate normal distribution model for clustering, exploiting asymptotic properties of Fourier coefficients.

Main Results:

• The proposed model-based clustering method demonstrated a lower error rate compared to K-means in simulations.
• The method achieved accurate clustering even with limited time points.
• Application to yeast cell cycle data yielded biologically interpretable gene groupings.

Conclusions:

• The developed statistical model effectively clusters genes based on similar expression change patterns.
• Derivative Fourier coefficients provide a powerful tool for analyzing temporal gene expression data.
• This approach offers a valuable method for gene classification and understanding biological dynamics.