DriverMEDS: Cancer driver gene identification using mutual exclusivity from embeded features and driver mutation...

Area of Science:

• Computational biology
• Genomics
• Cancer research

Background:

• Identifying cancer driver genes is crucial for cancer development, diagnosis, and treatment.
• Current methods often integrate multi-omics data and network embedding but overlook mutual exclusivity in the embedding space and assume high mutation frequencies for all driver genes.

Purpose of the Study:

• To develop an unsupervised framework, DriverMEDS, that leverages mutual exclusivity within learned gene features and optimizes mutation frequency scoring for improved cancer driver gene identification.

Main Methods:

• DriverMEDS utilizes a feature clustering algorithm to create gene modules, calculating module importance scores based on Euclidean distances of learned features and mutual exclusivity.
• A novel driver mutation scoring function is introduced, considering that most driver genes exhibit intermediate mutation frequencies.
• Gene prioritization is achieved through a weighted sum of module importance and driver mutation scores.

Main Results:

• DriverMEDS successfully detected novel cancer driver genes and relevant functional modules.
• Experimental analysis demonstrated that DriverMEDS outperforms five state-of-the-art methods for cancer driver identification.

Conclusions:

• DriverMEDS offers a more effective approach to unsupervised cancer driver gene identification by incorporating mutual exclusivity in the embedding space and refining mutation frequency analysis.
• The framework has the potential to advance cancer diagnosis and treatment strategies.