An agent-based learning model integrating sex differences in renal cell carcinoma

Area of Science:

• Computational oncology
• Immunogenomics
• Systems biology

Background:

• Sex-based differences significantly impact renal cell carcinoma (RCC) tumor biology, immune responses, and treatment outcomes.
• Existing computational models often fail to integrate sex hormones, immune composition, and tumor genetic evolution.
• Agent-based models (ABMs) are effective for simulating tumor-immune interactions but lack sex-specific modulation and machine learning optimization.

Purpose of the Study:

• To enhance an agent-based learning model (ALM) for simulating RCC progression and treatment response.
• To integrate hormonal effects, immune interactions, and tumor genetic adaptation into the ALM.
• To optimize the ALM using data-driven tuning for improved accuracy.

Main Methods:

• Developed an RCC-specific ALM incorporating immune agents (CD8+, NK, Treg, dendritic cells), hormone-sensitive mechanisms, and tumor genetic modules.
• Modeled tumor evolution using a genetic algorithm simulating mutations, with fitness based on immune evasion and proliferation advantages.
• Optimized model parameters using clinical outcomes from the ARON dataset via the Optuna framework, assessing performance with concordance index (CI) and mean squared error (MSE).

Main Results:

• Simulations successfully reproduced sex-specific treatment responses in renal cell carcinoma (RCC).
• Female models demonstrated delayed initial responses but stronger late immune activation and rapid tumor regression.
• Male models showed more stable early responses but exhibited greater tumor resilience due to genetic adaptations.

Conclusions:

• The enhanced ALM provides a framework for understanding the interplay of sex hormones, immune dynamics, and tumor genetics in RCC treatment outcomes.
• The model's adaptive learning capability reduced prediction error, suggesting potential for patient stratification.
• Further validation in larger cohorts is warranted to support the use of combined ABMs and data-driven optimization for patient prediction.