A Bayesian approach for accurate de novo transcriptome assembly.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• De novo transcriptome assembly from massive RNA-seq data presents challenges due to alternative splicing and variable gene expression.
• Existing assembly methods often produce inaccurate or mis-assembled transcripts, particularly for complex gene structures.

Purpose of the Study:

• To develop a robust and accurate de novo transcriptome assembler for RNA-seq data.
• To address the limitations of current assemblers in handling alternative splicing and expression levels.
• To identify phenotype-specific transcripts relevant to diseases like breast cancer.

Main Methods:

• Employs a read-guided strategy to reconstruct accurate splicing graphs from RNA-seq data.
• Utilizes a Bayesian strategy to estimate transcript expression probabilities without penalizing low-expression transcripts.
• Validated through simulation studies and benchmark tests on cell line data.

Main Results:

• BayesDenovo significantly reduces false positives and enhances assembly accuracy compared to other assemblers.
• Demonstrates superior performance for alternatively spliced genes and transcripts with varying expression levels.
• Shows increased robustness across multiple replicates, assembling a greater proportion of common transcripts.
• Identified phenotype-specific transcripts associated with breast cancer recurrence in clinical data.

Conclusions:

• BayesDenovo offers a highly accurate and robust solution for de novo transcriptome assembly.
• The method effectively handles challenges posed by alternative splicing and expression variability.
• Its application in cancer genomics can reveal novel disease-associated transcripts.