AI-Driver: an ensemble method for identifying driver mutations in personal cancer genomes

Area of Science:

• Genomics
• Computational Biology
• Cancer Research

Background:

• Distinguishing cancer driver mutations from passenger mutations is crucial for precision medicine.
• Existing methods for driver prediction lack the necessary resolution for exome sequencing data.

Purpose of the Study:

• To develop an advanced computational method for predicting the driver status of somatic missense mutations.
• To improve the accuracy and reliability of driver mutation identification in cancer research.

Main Methods:

• Developed an ensemble method named AI-Driver, utilizing 23 pathogenicity features.
• Trained and validated the AI-Driver model on extensive datasets, comparing its performance against individual tools and existing cancer-specific methods.
• Generated pre-computed AI-Driver scores for all human missense variants.

Main Results:

• AI-Driver demonstrated superior and stable performance across four independent benchmarks.
• The tool outperformed individual prediction tools and two existing cancer-specific driver prediction methods.
• Pre-computed scores are available for all human missense variants.

Conclusions:

• AI-Driver offers a significant advancement in identifying driver mutations from large-scale sequencing studies.
• The tool and its associated database are valuable resources for personal cancer genome analysis, drug target discovery, and biomarker development.
• AI-Driver supports precision cancer medicine by enabling more accurate identification of mutations driving tumor growth.