Freddie

Freddie detects annotation-independent alternative splicing isoforms from long-read (LR) transcriptomic sequencing to discover novel or aberrant splicing events relevant to cancer.

Key Features:

• Input Requirements: Accepts transcriptomic long-read (LR) sequencing data represented as genomic alignments produced by a splice aligner.
• Read Partitioning: Partitions reads into independent sets for parallel processing.
• Canonical Exon Segmentation: Segments genomic alignments into canonical exon segments representing potential isoforms, formulated and solved as an optimization problem using Dynamic Programming.
• Isoform Reconstruction: Reconstructs isoforms by jointly clustering and error-correcting reads using the canonical segmentation and solves the Minimum Error Clustering into Isoforms (MErCi) problem via Integer Linear Programming (ILP).
• Performance Evaluation: Shows improved recall over other isoform detection tools on simulated datasets while maintaining a comparable false positive rate.
• Practical Application: Detected a potentially novel Androgen Receptor isoform characterized by intron retention in a prostate cancer cell line using LR data and cross-validated the finding with short-read RNA-seq.

Scientific Applications:

• Cancer research: Enables detection and discovery of novel or aberrant alternative splicing events to investigate molecular mechanisms and progression in cancer.
• Biomarker and therapeutic target discovery: Identifies candidate cancer-specific splicing events, including novel isoforms, for potential personalized therapeutic strategies.

Methodology:

Uses genomic alignments from a splice aligner, partitions reads, segments alignments into canonical exon segments via an optimization solved with Dynamic Programming, and reconstructs isoforms by joint clustering/error-correction solving the MErCi problem with Integer Linear Programming (ILP).