FusionGDB 2.0

FusionGDB 2.0 provides systematic functional annotation of human fusion genes, integrating deep learning-derived breakage tendency scoring, transcribed chimeric sequence and open reading frame analyses, genomic feature overlap assessments, and protein feature retention evaluation to support interpretation of fusion events.

Key Features:

• Up-to-Date Human Fusion Genes: FusionGDB 2.0 contains a comprehensive collection of current human fusion genes for annotation and analysis.
• Fusion Gene Breakage Tendency Score: Uses the FusionAI deep learning model to assess breakage tendencies from DNA sequences surrounding breakpoints within 20 kb regions.
• Genomic Feature Overlap Analysis: Investigates overlaps between fusion breakpoints and 44 human genomic features categorized across five cellular roles.
• Transcribed Chimeric Sequence and ORF Analysis: Analyzes transcribed chimeric sequences and open reading frames using deep learning approaches that incorporate Ribo-seq read features to evaluate coding potential.
• Protein Feature Retention Investigation: Examines retention of protein features from individual fusion partner genes at the protein level.
• Enhanced ORF Annotation: Annotates ORF frame status across ~102,000 fusion genes, with approximately 15,000 maintained as in-frame ORFs.
• Annotation Categories: Provides eight categories of functional annotations for fusion genes.

Scientific Applications:

• Reference knowledgebase for genomic studies: Serves as a curated resource for analysis and interpretation of human fusion genes.
• Cancer genomics: Supports investigation of fusion-driven oncogenesis and genomic breakage contexts.
• Molecular diagnostics: Enables identification and characterization of fusion-derived biomarkers for diagnostic assays.
• Therapeutic development: Informs discovery and prioritization of fusion-targeted therapies and biomarkers.
• Evaluation of coding potential: Assists assessment of coding potential of chimeric transcripts using Ribo-seq-informed analyses.

Methodology:

Applied the FusionAI deep learning model to DNA sequences surrounding breakpoints within 20 kb to compute breakage tendency scores; computed overlaps between breakpoints and 44 genomic features across five cellular roles; performed deep learning analyses of transcribed chimeric sequences and ORFs incorporating Ribo-seq read features; examined protein-level retention of partner protein features; and annotated ORF frame status for ~102,000 fusion genes (≈15,000 in-frame).