Freddie: annotation-independent detection and discovery of transcriptomic alternative splicing isoforms using...

Area of Science:

• Genomics
• Bioinformatics
• Cancer Research

Background:

• Alternative splicing (AS) is crucial in cancer development and a potential therapeutic target.
• Detecting novel AS events is challenging due to incomplete databases and limitations of short-read sequencing.
• Long-read (LR) sequencing offers potential for improved AS detection and discovery.

Purpose of the Study:

• To introduce Freddie, a computational tool for annotation-independent isoform discovery and detection using LR sequencing.
• To address the limitations of current AS detection methods, particularly for novel and cancer-specific events.
• To enable more comprehensive analysis of transcriptomes for cancer research.

Main Methods:

• Freddie utilizes transcriptomic LR sequencing and genomic split alignments.
• It employs dynamic programming for canonical exon segmentation and integer linear programming (ILP) for isoform clustering and error correction (MErCi problem).
• The tool processes reads in parallel partitions for efficiency.

Main Results:

• Freddie demonstrated superior accuracy in detecting AS events on simulated datasets compared to other tools.
• It outperformed existing methods even when provided with complete ground truth annotations.
• Analysis of a prostate cancer cell line dataset showed Freddie identified isoforms with higher short-read cross-validation rates.

Conclusions:

• Freddie is an effective tool for discovering and detecting alternative splicing isoforms from LR sequencing data.
• Its annotation-independent approach overcomes limitations of current databases and methods.
• Freddie advances cancer research by enabling more accurate identification of AS events for potential therapeutic targeting.