EnFusion

EnFusion detects gene fusion events by integrating multiple fusion-calling algorithms to improve sensitivity and specificity for identifying somatic gene fusions in RNA-Seq data from pediatric cancers.

Key Features:

• Ensemble Approach: Integrates outputs from Arriba, CICERO, FusionMap, FusionCatcher, JAFFA, MapSplice, and STAR-Fusion to generate consensus fusion calls.
• Implementation: Implemented as an automated pipeline using Docker and Amazon Web Services (AWS) serverless technology.
• Consensus Detection: Calls fusion events when at least three algorithms agree, increasing confidence and reducing algorithm-specific false positives.
• Knowledge-Based Filtering: Compares consensus results against an internal cohort-specific database of artifactual fusions and applies a "known fusion list" to retain clinically relevant and pathogenic events.
• Clinical Dataset Application: Applied to RNA-Seq data from 229 pediatric cancer patients, including central nervous system tumors, solid tumors, and hematologic malignancies or disorders, with a reported diagnostic yield of 29.3% and detected fusions such as RBPMS-MET, BCAN-NTRK1, and TRIM22-BRAF.
• Novel and Known Fusion Identification: Detects both known and novel fusion events that can influence diagnosis and treatment decision-making and suggest potential targeted therapies.

Scientific Applications:

• Pediatric oncology diagnostics: Detection of somatic driver fusions to inform diagnosis and patient management in pediatric central nervous system, solid, and hematologic malignancies.
• Fusion discovery and characterization: Identification and curation of novel and clinically significant fusions for downstream validation and therapeutic consideration.
• Molecular oncology research: Elucidation of fusion-driven tumorigenic mechanisms and contribution to studies of cancer genomic landscapes distinct from adult malignancies.

Methodology:

Integrates outputs from Arriba, CICERO, FusionMap, FusionCatcher, JAFFA, MapSplice, and STAR-Fusion; requires consensus of ≥3 algorithms for fusion calls; filters consensus results using an internal cohort-specific database and a known fusion list; implemented with Docker and AWS serverless and applied to RNA-Seq data from pediatric cohorts.