Machine Learning Methods for Precision Dosing in Anticancer Drug Therapy: A Scoping Review.

Area of Science:

• Oncology
• Artificial Intelligence
• Pharmacology

Background:

• Machine Learning (ML) techniques have emerged for individualizing anticancer drug doses.
• Approaches often rely on presumed drug effects or measured biomarkers.
• Precision dosing in cancer therapy is a key area of research.

Purpose of the Study:

• To comprehensively summarize the research status of ML in precision dosing for anticancer drugs.
• To review studies published after 2016.

Main Methods:

• A scoping review was conducted following Cochrane and Joanna Briggs Institute guidelines.
• Systematic searches of Medline, Embase, and Cochrane Library databases were performed.
• Reporting adhered to the PRISMA-ScR checklist.

Main Results:

• 17 relevant studies were identified.
• Reinforcement Learning (RL) methods were used in 12 studies, including various Q-Learning algorithms.
• Classical ML methods were compared, and an AI platform guided dosing in limited patients.

Conclusions:

• ML methods are clinically relevant for optimizing anticancer drug doses.
• Many ML algorithms demonstrate potential for model-free predictions.
• ML approaches may maximize efficacy and minimize toxicity compared to standard protocols.