Ryūtō: network-flow based transcriptome reconstruction.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• High-throughput RNA sequencing (RNA-seq) has advanced transcript detection but faces challenges in complete genome annotation.
• Computational reconstruction of transcript structures is a bottleneck due to data noise and biases.
• Accurate annotation of complex transcriptional output remains elusive.

Purpose of the Study:

• To develop and present a novel computational workflow for improved transcript assembly and quantification.
• To address the limitations in current RNA-seq data analysis for accurate transcript structure reconstruction.
• To enhance the detection of both coding and non-coding RNA transcripts.

Main Methods:

• Introduced new and improved algorithms within a unified workflow for transcript assembly and quantification.
• Extended the splice graph framework by integrating overlap and bin graphs.
• Utilized phasing information of reads and modeled read coverage decomposition as a minimum-cost flow problem.

Main Results:

• The novel workflow efficiently utilizes multi-splice and paired-end RNA-seq data.
• Phasing information was effectively employed to resolve complex loci.
• The minimum-cost flow model addressed non-uniformities inherent in RNA-seq data.

Conclusions:

• The developed workflow demonstrates superior performance compared to state-of-the-art methods on simulated and real datasets.
• Ryūtō identified 1-4% more true transcripts than competing methods.
• Ryūtō achieved a 5-35% reduction in false predictions compared to the next best competitor.