Network as a Biomarker: A Novel Network-Based Sparse Bayesian Machine for Pathway-Driven Drug Response Prediction.

Area of Science:

• Systems biology
• Bioinformatics
• Computational oncology

Background:

• Advances in biological networks and analysis enable molecular biomarker discovery for cancer treatment monitoring.
• Disease-specific driver signaling networks have been previously reconstructed using multi-omics and pathway data.

Purpose of the Study:

• To develop a network-based sparse Bayesian machine (NBSBM) for predicting cancer cell drug responses.
• To utilize disease-specific driver signaling networks to enhance prediction accuracy in high-dimensional, low-data scenarios.

Main Methods:

• Developed a network-based sparse Bayesian machine (NBSBM) incorporating spike and slab prior distributions and Markov random field priors.
• Applied NBSBM to previously reconstructed disease-specific driver signaling networks.
• Compared NBSBM performance against network-based support vector machine (NBSVM) methods.

Main Results:

• NBSBM achieved significantly higher accuracy in predicting cancer cell drug responses compared to NBSVM methods.
• Identified gene modules from disease-specific networks potentially involved in drug sensitivity or resistance.
• Demonstrated improved prediction performance with reduced training data and high-dimensional feature spaces.

Conclusions:

• The NBSBM approach provides a robust method for disease-specific network-based drug sensitivity prediction.
• This method can uncover potential drug mechanisms of action by identifying predictive sub-networks.
• The findings support the advancement of biomarker-driven personalized medicine in cancer treatment.