Agent-Based Simulations of Lung Tumor Evolution Suggest That Ongoing Cell Competition Drives Realistic Clonal...

Area of Science:

• Oncology
• Computational Biology
• Genetics

Background:

• Computational simulations are vital for understanding tumor evolution and cancer growth dynamics.
• Accurate tumor models are essential for reliable inferences about cancer development.
• Lung tumors exhibit frequent, late subclonal expansions linked to poor prognosis, necessitating models that capture this behavior.

Purpose of the Study:

• To evaluate three computational models of 3D tumor growth for their ability to replicate late subclonal expansions observed in lung tumors.
• To identify a computationally efficient model that generates realistic multi-region sequencing data for lung cancers.
• To assess the impact of model assumptions on inferring driver mutation fitness effects in lung cancer.

Main Methods:

• Tested three distinct 3D tumor growth simulation models with varying cell competition assumptions.
• Validated model performance against known properties of lung tumor evolution, specifically late subclonal expansions.
• Applied inference pipelines to a large lung cancer cohort using both single-stage and the identified two-stage models.

Main Results:

• Identified a computationally efficient two-stage growth model that accurately simulates lung tumor evolution and generates realistic sequencing data.
• The two-stage model incorporates local competition within fixed-size tumors after an initial growth phase.
• Inference pipelines using the two-stage model suggested significantly larger selection effects for driver mutations compared to single-stage models.

Conclusions:

• Model assumptions critically influence the outcomes of tumor-specific, simulation-based inferences.
• A two-stage tumor growth model with local competition provides a more realistic representation of lung cancer evolution.
• The choice of simulation model has substantial implications for understanding the impact of driver mutations in cancer.