AI-based neoadjuvant immunotherapy response prediction across pan-cancer: a comprehensive review

Area of Science:

• Oncology
• Artificial Intelligence
• Biomedical Informatics

Background:

• Neoadjuvant immunotherapy (NIT) shows promise but patient response varies significantly due to tumor heterogeneity.
• Predicting NIT response preoperatively is crucial for personalized cancer treatment.
• Artificial intelligence (AI), including machine learning (ML) and deep learning (DL), offers new methods for response prediction.

Purpose of the Study:

• To systematically review AI-driven computational approaches for predicting neoadjuvant immunotherapy response.
• To categorize AI models based on predictive paradigms (indirect vs. direct) and data modalities (radiomics, pathomics, genomics, multi-omics).
• To identify current challenges and discuss strategies for advancing AI in precision immunotherapy.

Main Methods:

• Systematic review of AI-driven computational approaches for NIT response prediction.
• Categorization of AI models into indirect (surrogate biomarkers) and direct (data-driven biomarkers) paradigms.
• Classification of models based on data modalities: radiomics, pathomics, genomics, and multi-omics.

Main Results:

• AI approaches can extract features from high-dimensional data to build predictive models for NIT response.
• Indirect paradigm uses AI to predict surrogate biomarkers, while the direct paradigm uses AI to identify data-driven biomarkers for clinical endpoints.
• Various data modalities offer distinct insights into tumor characteristics and treatment response.

Conclusions:

• AI holds significant potential for optimizing individualized treatment strategies in oncology.
• Current AI predictive models face biomarker-based and AI-based challenges that require further research.
• Accelerating AI integration into precision immunotherapy requires addressing these limitations and guiding future research.