Model-based clustering for RNA-seq data.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• RNA-sequencing (RNA-seq) is a powerful tool for global gene expression analysis, surpassing microarrays.
• Existing statistical tools for RNA-seq data analysis are limited, particularly for complex datasets.
• Cluster analysis is crucial for understanding gene functions and networks from RNA-seq data.

Purpose of the Study:

• To develop robust, probability-based clustering algorithms for RNA-seq data.
• To enhance the accuracy and efficiency of gene expression data analysis.
• To provide tools for visualizing gene relationships and selecting optimal cluster numbers.

Main Methods:

• Derivation of expectation-maximization (EM) and stochastic EM algorithms tailored for RNA-seq data.
• Implementation of a likelihood-based initialization strategy to improve clustering performance.
• Development of a hybrid-hierarchical clustering method for visualizing cluster relationships.

Main Results:

• Proposed algorithms demonstrated superior clustering performance over K-means and non-model-based hierarchical methods.
• Simulation studies confirmed the effectiveness of the developed methods.
• Analysis of a maize RNA-seq dataset validated the practical utility and accuracy of the algorithms.

Conclusions:

• The developed probability-based clustering algorithms significantly advance RNA-seq data analysis.
• The MBCluster.Seq R package provides an efficient and accessible tool for researchers.
• These methods offer improved insights into gene functions and biological networks from RNA-seq experiments.