Cross-Scale Sensitivity Analysis of a Non-Small Cell Lung Cancer Model: Linking Molecular Signaling Properties to Cellular Behavior

Sensitivity analysis is an effective tool for systematically identifying specific perturbations in parameters that have significant effects on the behavior of a given biosystem, at the scale investigated. In this work, using a two-dimensional, multiscale non-small cell lung cancer (NSCLC) model, we examine the effects of perturbations in system parameters which span both molecular and cellular levels, i.e. across scales of interest. This is achieved by first linking molecular and cellular activities and then assessing the influence of parameters at the molecular level on the tumor’s spatio-temporal expansion rate, which serves as the output behavior at the cellular level. Overall, the algorithm operated reliably over relatively large variations of most parameters, hence confirming the robustness of the model. However, three pathway components (proteins PKC, MEK, and ERK) and eleven reaction steps were determined to be of critical importance by employing a sensitivity coefficient as an evaluation index. Each of these sensitive parameters exhibited a similar changing pattern in that a relatively larger increase or decrease in its value resulted in a lesser influence on the system’s cellular performance. This study provides a novel cross-scaled approach to analyzing sensitivities of computational model parameters and proposes its application to interdisciplinary biomarker studies.

💡 Research Summary

The paper presents a novel cross‑scale sensitivity analysis applied to a two‑dimensional multiscale model of non‑small cell lung cancer (NSCLC). The authors integrate intracellular molecular signaling—specifically the EGFR‑RAS‑RAF‑MEK‑ERK cascade—with extracellular cellular dynamics such as proliferation, migration, and death on a spatial grid. Each biochemical reaction is described by ordinary differential equations, while the cell population evolves according to agent‑based rules that generate a spatio‑temporal tumor expansion pattern. The primary output of interest is the tumor’s expansion rate, which serves as a measurable cellular‑level phenotype.

To quantify how variations in molecular parameters affect this phenotype, the authors define a sensitivity coefficient S = (ΔO/O)/(ΔP/P), where O is the tumor expansion rate and P is a model parameter (e.g., reaction rate constant, initial concentration). They systematically perturb each of roughly thirty parameters across a wide range (−200 % to +200 % of the baseline) and compute S for each perturbation. Most parameters exhibit low sensitivity (|S| < 0.1), indicating that the model is robust to moderate fluctuations and that the tumor growth phenotype is buffered against many molecular changes.

However, three proteins—PKC, MEK, and ERK—along with eleven specific reaction steps, emerge as highly sensitive (|S| > 0.5). In particular, the MEK→ERK activation step shows the strongest influence on tumor expansion. Notably, the relationship between parameter magnitude and sensitivity is non‑linear: as a parameter is increased or decreased far beyond its baseline, the corresponding S value diminishes, suggesting saturation or feedback inhibition within the signaling network. This behavior reflects biological reality, where downstream pathways often become saturated or are regulated by negative feedback loops, limiting the impact of extreme molecular perturbations.

The authors cross‑validate their computational findings with experimental and clinical literature. The identified importance of MEK and ERK aligns with numerous studies showing that MEK or ERK inhibitors can suppress NSCLC growth. The sensitivity of PKC, while less explored clinically, points to a potential therapeutic target that warrants further investigation. By demonstrating that a small subset of molecular parameters disproportionately governs the macroscopic tumor behavior, the study provides a concrete framework for prioritizing biomarkers and drug targets.

Beyond the specific NSCLC context, the paper proposes that this cross‑scale sensitivity methodology can be generalized to other complex biological systems. It offers a systematic way to bridge molecular perturbations and phenotypic outcomes, facilitating interdisciplinary biomarker discovery, rational drug design, and personalized medicine strategies. The robustness of the model under large parameter variations, coupled with the clear identification of critical nodes, underscores the utility of integrating multiscale modeling with rigorous sensitivity analysis in cancer research.