Reinforcement Learning for Precision Oncology.

Area of Science:

• Oncology
• Machine Learning
• Computational Biology

Background:

• Precision oncology utilizes genetic insights for tailored cancer treatments.
• Machine learning, including supervised and unsupervised methods, aids risk assessment and outcome prediction.
• Reinforcement learning (RL), though less explored, offers potential for optimizing sequential treatment decisions.

Purpose of the Study:

• To review recent advances in reinforcement learning (RL) applications in oncology.
• To identify and discuss challenges and pitfalls in developing RL-based decision support systems for precision oncology.

Main Methods:

• Review of recent studies applying reinforcement learning in oncology.
• Analysis of challenges related to applicability, validity, and safety of RL in clinical settings.

Main Results:

• Reinforcement learning has demonstrated success in other complex sequential tasks.
• RL holds significant potential as a decision support tool in precision oncology.
• Several challenges must be overcome for successful clinical implementation of RL.

Conclusions:

• Reinforcement learning is a promising, yet underutilized, tool for advancing precision oncology.
• Future research must address identified challenges to ensure the safe and effective development of RL-based clinical decision support systems.