A clustering procedure for three-way RNA sequencing data using data transformations and matrix-variate Gaussian...

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• RNA sequencing (RNA-seq) time-course experiments generate complex three-way count data (genes, time points, biological units).
• Clustering RNA-seq data is crucial for identifying co-expressed genes over time.
• Normalized RNA-seq counts exhibit compositional data properties, necessitating specialized analytical methods.

Purpose of the Study:

• To develop and validate a robust procedure for clustering three-way RNA-seq count data.
• To effectively analyze gene expression profiles across time and biological variations.
• To improve the extraction of biologically relevant gene expression patterns.

Main Methods:

• Pre-processing RNA-seq data to obtain normalized expression profiles.
• Applying the additive log ratio (CLR) transform to convert compositional data to Euclidean vectors.
• Utilizing matrix-variate Gaussian mixture models for clustering the transformed data.
• Assessing cluster quality using density-based silhouette information and cluster maps for visualization.

Main Results:

• The proposed method successfully clusters three-way RNA-seq data, considering its compositional nature.
• Evaluation metrics demonstrate effective cluster separation and compactness.
• The procedure was illustrated using RNA-seq data from fission yeast, showing comparable results to a two-way approach.

Conclusions:

• The developed procedure offers a suitable approach for clustering three-way RNA-seq data.
• Compositional data analysis techniques, like the CLR transform, are beneficial for RNA-seq data.
• This method enhances the understanding of gene co-expression dynamics in complex experimental designs.