Artificial intelligence and machine learning methods in predicting anti-cancer drug combination effects

Area of Science:

• Computational Biology
• Pharmacology
• Artificial Intelligence in Medicine

Background:

• Drug combinations show promise for cancer therapy with reduced toxicity.
• Experimental screening of all possible drug combinations is infeasible due to vast search space.
• Computational models are crucial for predicting synergistic anti-cancer drug combinations.

Purpose of the Study:

• To provide a structured overview of machine learning (ML), especially deep learning (DL), methods for anti-cancer drug synergy prediction.
• To present a unified framework for ML models, detailing architectures, contributions, and limitations.
• To compare the prediction performance of reviewed computational models through unbiased experiments.

Main Methods:

• Review of large-scale databases relevant to drug synergy prediction.
• Systematic analysis of machine learning and deep learning model architectures.
• Development of a unified framework for ML-based predictive models.
• Conducting comparative experiments to evaluate prediction performance.

Main Results:

• Machine learning and deep learning methods are increasingly popular and achieve state-of-the-art performance in predicting anti-cancer drug synergy.
• These data-driven methods offer advantages over traditional hypothesis-driven approaches by requiring fewer mechanistic assumptions.
• Comparative experiments provide insights into the relative strengths and weaknesses of different computational models.

Conclusions:

• Machine learning, particularly deep learning, represents a powerful and efficient approach for identifying synergistic anti-cancer drug combinations.
• The structured overview and comparative analysis guide future development of more accurate and reliable computational predictive models.
• Advancements in this field hold significant potential for optimizing cancer treatment strategies.