Implementing generative artificial intelligence in precision oncology: safety, governance, and significance

Area of Science:

• Oncology
• Artificial Intelligence
• Bioinformatics

Background:

• Precision oncology faces challenges in improving patient outcomes and requires advanced tools for data integration.
• Generative AI has rapidly evolved, offering potential solutions for efficiency and adoption in healthcare.

Purpose of the Study:

• To review the development and healthcare applications of generative AI in precision oncology.
• To outline a strategy for implementing generative AI to enhance cancer care, focusing on mutation interpretation, trial matching, and tumor phenotype analysis.

Main Methods:

• Surveying generative AI's history and current healthcare uses.
• Proposing a strategy using retrieval-augmented generation (RAG) and human-in-the-loop (HITL) workflows.
• Incorporating data standards (OMOP, mCODE, FHIR), prospective evaluation, and synthetic data generation.

Main Results:

• Generative AI can aid in interpreting genetic mutations and their pathological significance.
• AI-driven verification can streamline clinical trial eligibility matching.
• Multimodal foundation models can compute tumor phenotypes from real-world data for report drafting and molecular surrogate estimation.

Conclusions:

• Generative AI offers significant potential to advance precision oncology by improving standardization, cost-effectiveness, and data harmonization.
• A proposed RAG and HITL strategy, with robust data governance and synthetic data, can accelerate patient-centered decision-making and clinical trial development.
• This approach paves the way for a