Comprehensive evaluation of computational methods for predicting cancer driver genes.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Identifying cancer driver genes is essential for understanding tumorigenesis and developing targeted therapies.
• Numerous computational methods exist, including frequency, network, and function-based approaches, but a comprehensive performance evaluation is lacking.

Purpose of the Study:

• To comprehensively evaluate and compare the performance of various computational methods for identifying cancer driver genes.
• To identify the most accurate and reliable algorithms for predicting candidate driver genes.

Main Methods:

• Assessed eight performance criteria for eight network-based, one function-based, and three frequency-based algorithms.
• Utilized eight benchmark datasets to compare algorithm performance under different conditions (network type, measurement, sample size).

Main Results:

• Performance varied across methods depending on the dataset and criteria.
• Frequency-based driverMAPS and network-based HotNet2 demonstrated the best overall performance.
• Network-based algorithms utilizing protein-protein interaction networks generally outperformed function- and frequency-based approaches.

Conclusions:

• Most current algorithms require stringent cutoffs to accurately distinguish driver from non-driver genes.
• The study developed the Cancer Driver Catalog website, integrating gene scores from various software to guide researchers in prioritizing cancer driver gene candidates.