Explainable Artificial Intelligence Reveals Novel Insight into Tumor Microenvironment Conditions Linked with Better...

Area of Science:

• Immunology
• Oncology
• Artificial Intelligence

Background:

• The tumor microenvironment (TME) plays a crucial role in breast cancer progression and patient outcomes.
• Understanding the relationship between immune cell composition and long-term survival is vital for improving breast cancer treatment.

Purpose of the Study:

• To investigate the data-driven relationship between immune cell composition in the TME and ≥5-year survival rates in breast cancer patients.
• To utilize explainable artificial intelligence (XAI) models to identify critical immune cell types and their prognostic significance.

Main Methods:

• Acquired TCGA breast invasive carcinoma data.
• Retrieved immune cell composition estimates using TIMER2.0 (EPIC, CIBERSORT, TIMER, xCell).
• Applied XAI models to analyze TME features and survival data.

Main Results:

• XAI models identified B cells, CD8+ T cells, M0 macrophages, and NK T cells as critical for enhanced prognosis.
• Determined specific inflection points for these immune cells associated with improved ≥5-year survival.
• Quantified conditional probabilities of survival based on immune cell fractions: B cell >0.025, M0 macrophage <0.05, NK T cell >0.075, CD8+ T cell >0.25.

Conclusions:

• Specific immune cell compositions in the TME are strongly predictive of ≥5-year survival in breast cancer patients.
• Findings support the potential for XAI-driven clinical predictions and enhanced immunotherapies.
• Identified thresholds for immune cell fractions offer targets for TME reprogramming strategies.