GPS-Net: discovering prognostic pathway modules based on network regularized kernel learning.

Area of Science:

• Computational biology
• Genomics
• Bioinformatics

Background:

• Current prognostic biomarker discovery relies heavily on single-gene or global-gene approaches.
• These methods often overlook complex gene interactions and pathway alterations crucial for understanding diseases like cancer.

Purpose of the Study:

• To introduce GPS-Net, a novel computational framework for identifying prognostic gene modules.
• To improve the accuracy and efficiency of prognostic biomarker and pathway identification in complex diseases.

Main Methods:

• Developed GPS-Net, a framework integrating holistic pathway structures and gene interaction networks.
• Employed advanced multiple kernel learning and network-based regularization techniques.
• Validated through extensive simulation studies.

Main Results:

• GPS-Net enhances the accuracy of biomarker and pathway identification.
• Significantly reduces computational complexity compared to existing methods.
• Identified key predictive pathways for patient outcomes in a cancer immunotherapy study.

Conclusions:

• GPS-Net offers a scalable and feasible framework for genome-wide, pathway-level prognostic analysis.
• Synergizes mechanism-driven and data-driven approaches for precision genomics.
• Provides a powerful tool for understanding complex diseases and improving patient outcomes.