LOTUS

LOTUS predicts cancer driver genes by integrating gene mutation and protein-protein interaction data with machine learning to identify oncogenes and tumor suppressor genes across cancer types.

Key Features:

• Data integration: Integrates gene mutation and protein-protein interaction data using machine learning to inform driver gene prediction.
• Multitask learning: Implements a multitask learning strategy to share information across different cancer types.
• Prediction scope: Produces both pan-cancer predictions and cancer-type-specific analyses.
• Driver types: Predicts oncogenes and tumor suppressor genes as candidate cancer drivers.
• Framework: Uses a versatile framework to enhance prediction accuracy across diverse contexts.
• Benchmarking: Demonstrates superiority over five other driver gene prediction methods in intrinsic consistency and prediction accuracy, including identification of novel cancer genes across multiple cancer types.

Scientific Applications:

• Driver gene identification: Prioritizes candidate cancer driver genes for use as potential therapeutic targets or biomarkers.
• Pan-cancer discovery: Identifies novel cancer genes across multiple cancer types.
• Tumor biology insight: Supports analysis of mechanisms that confer selective growth advantages, including uncontrolled proliferation and evasion of apoptosis.

Methodology:

Applies machine learning to integrate gene mutation and protein-protein interaction data and employs a multitask learning strategy for pan-cancer and cancer-type-specific predictions, with performance evaluated by intrinsic consistency and prediction accuracy against five other methods.