A Genetic Algorithm-Based Ensemble Learning Framework for Drug Combination Prediction

Area of Science:

• Computational biology
• Bioinformatics
• Machine learning in drug discovery

Background:

• Combination therapy offers improved efficacy and reduced resistance for complex diseases like cancer.
• Predicting synergistic drug combinations is crucial but challenged by imbalanced datasets where synergistic pairs are rare.
• Existing prediction models struggle with class imbalance and high-dimensional biological data.

Purpose of the Study:

• To develop an effective computational framework for predicting synergistic drug combinations across various cancer cell lines.
• To address the challenges of class imbalance and high dimensionality inherent in drug combination datasets.
• To improve the identification of clinically relevant synergistic drug combinations.

Main Methods:

• Proposed GA-DRUG, a genetic algorithm-based ensemble learning framework.
• Utilized cell-line-specific gene expression profiles under drug perturbations for model training.
• Incorporated imbalanced data processing and global optimal solution search mechanisms.

Main Results:

• GA-DRUG outperformed 11 state-of-the-art algorithms in predicting synergistic drug combinations.
• Demonstrated significant improvement in predicting the minority class (Synergy).
• Experimental validation using cellular proliferation assays confirmed GA-DRUG's predictive accuracy.

Conclusions:

• GA-DRUG provides a robust solution for predicting synergistic drug combinations, particularly for rare synergistic events.
• The ensemble framework effectively corrects individual classifier errors, enhancing overall prediction performance.
• This approach holds promise for accelerating the discovery of effective combination therapies in oncology.