Polee: RNA-Seq analysis using approximate likelihood.

Area of Science:

• Bioinformatics
• Computational Biology
• Molecular Biology

Background:

• RNA sequencing (RNA-Seq) is crucial for quantifying mRNA transcript abundance.
• Current RNA-Seq analyses often neglect estimation uncertainty, particularly for complex scenarios like isoforms and duplicated genes.
• Full probabilistic models for RNA-Seq data are computationally intractable due to massive datasets.

Purpose of the Study:

• To develop a computationally efficient method for RNA-Seq analysis that accounts for uncertainty.
• To improve the accuracy of detecting differential transcript expression and coexpression.
• To address the limitations of existing RNA-Seq analysis methods regarding uncertainty and computational cost.

Main Methods:

• Developed a novel approximation for the likelihood function of a sparse mixture model.
• Introduced the Pólya tree transformation technique for this approximation.
• Evaluated the method's performance in detecting differential and coexpressed transcripts.

Main Results:

• The Pólya tree transformation approximation retains most of the benefits of full probabilistic models.
• The proposed method significantly reduces computational costs compared to existing approaches.
• Achieved more accurate detection of differential transcript expression and transcript coexpression.

Conclusions:

• The Pólya tree transformation offers a computationally feasible solution for incorporating uncertainty into RNA-Seq analysis.
• This method enhances the reliability of differential and coexpression analyses.
• The approach provides a more accurate and efficient tool for molecular biology research.