sysSVM2

sysSVM2 prioritizes patient-specific cancer driver genes by integrating molecular alterations and gene systems-level properties to support precision oncology.

Key Features:

• Patient-Specific Driver Gene Prioritization: Prioritizes cancer driver genes for individual patients using molecular and systems-level gene properties.
• Molecular Features: Incorporates mutations and copy-number variants as molecular inputs for driver prediction.
• Systems-Level Features: Utilizes systems-level properties including evolutionary origin and expression breadth as predictive features.
• One-Class SVM Framework: Employs a one-class Support Vector Machine to identify genes with characteristics similar to known canonical drivers.
• Integration of Genetic and Systems-Level Data: Integrates cancer genetic alterations with gene systems-level properties to produce patient-specific driver predictions.
• Optimization and Benchmarking: Optimized on simulated pan-cancer data and benchmarked on real cancer datasets with reported low false-positive rate.
• Applicability to Rare Cancer Types: Applicable to rare cancers with limited known driver genes and yields predicted drivers that disrupt known cancer-related pathways.
• Models Trained on TCGA Data: Includes models trained on The Cancer Genome Atlas (TCGA) data.

Scientific Applications:

• Precision Oncology: Identifies patient-specific driver genes to inform personalized therapeutic decision-making.
• Research in Rare Cancers: Enables driver discovery in cancer types with limited cohort-level recurrence of canonical drivers.
• Pathway Disruption Analysis: Facilitates analysis of how predicted driver genes disrupt known cancer-related pathways.

Methodology:

Features combining molecular alterations (mutations, copy-number variants) and systems-level properties (evolutionary origin, expression breadth) are input to a one-class Support Vector Machine; models were optimized on simulated pan-cancer data, benchmarked on real cancer datasets, and include models trained on TCGA data.