ArtiFuse

ArtiFuse simulates fusion genes by modifying genomic reference sequences to enable realistic benchmarking of RNA-seq fusion detection tools.

Key Features:

• Reference sequence modification: Simulates fusion events by directly editing the genomic reference at specified breakpoints.
• Real RNA-seq benchmarking: Applies simulated fusions to real RNA-seq datasets without requiring simulated reads.
• Customizable simulations: Allows specification of involved genes, breakpoint positions, and control over expression levels.
• Performance evaluation: Computes recall values for existing fusion detection tools across datasets, with reported average recall peaking between 0.4 and 0.56 for high-quality, high-coverage RNA-seq.
• Gene property analysis: Assesses detection performance in relation to gene expression levels and co-expression with paralogues, identifying decreased recall for low-expressed genes and genes with co-expressed paralogues.

Scientific Applications:

• Benchmarking fusion detection tools: Provides realistic evaluation datasets for comparing sensitivity of RNA-seq fusion callers.
• Algorithm development: Supports development and refinement of fusion gene prediction algorithms by revealing failure modes related to expression and paralogue co-expression.
• Cancer research and diagnostics assessment: Informs assessment of fusion detection applicability in cancer studies and diagnostic contexts using real-sample RNA-seq data.

Methodology:

Simulates fusion genes by direct sequence modification of the genomic reference at specific breakpoints, applies these modifications to real RNA-seq datasets without generating simulated reads, and controls involved genes and their expression levels.