Efficient RNA isoform identification and quantification from RNA-Seq data with network flows.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Current isoform identification methods, often using L1-regularized regression (Lasso), struggle with genes containing numerous exons due to the intractability of listing all possible transcripts.
• Existing L1-penalty approaches are limited to genes with few exons or rely on preselected transcript sets, potentially missing important isoforms.

Purpose of the Study:

• To develop a novel computational technique for accurate and efficient isoform identification and quantification.
• To overcome the limitations of existing methods by handling the full set of candidate transcripts without preselection.

Main Methods:

• Introduction of FlipFlop, a new technique employing network flow optimization to address sparse estimation problems.
• FlipFlop efficiently processes the complete set of candidate isoforms, eliminating the need for a preliminary isoform selection step.

Main Results:

• FlipFlop demonstrates superior accuracy in isoform identification compared to alternative methods.
• The technique maintains low computational cost, offering one of the fastest available solutions.
• Experiments with both synthetic and real RNA-Seq data validate the method's performance.

Conclusions:

• FlipFlop provides a significant advancement in isoform identification and quantification, particularly for complex genes.
• The method's efficiency and accuracy make it a valuable tool for RNA-Seq data analysis.
• The freely available R package facilitates broader adoption and application in the research community.