Computational models reveal shared dynamics in lung tumors and TB granulomas

Both small cell lung cancer (SCLC) and tuberculosis (TB) form complex, uneven lesions in the lung that are hard to study directly in people. These lesions often include tight interactions between T cells and macrophages, and patients show wide differences in how disease progresses and how treatments work. Denise Kirschner and colleagues set out to study those differences using computer-based experiments instead of only lab or clinical studies. The team has previously built GranSim, an agent-based model that simulates how TB granulomas form, grow and respond to treatment at the scale of individual cells. In the new work they created TumorSim, an analogous agent-based model designed to capture how SCLC tumors interact with the lung immune environment over space and time. By building TumorSim on known lung immunology and on literature about lung tumors, the researchers aimed to bridge gaps that traditional experiments struggle to address: how immune cells arrive, move and change roles inside a growing lesion, and why otherwise similar lesions can end up very differently.

TumorSim is a mechanistic, spatio-temporal model that encodes many immune-tumor interactions described in the literature. The model includes cytokine-based recruitment of adaptive cells and the PD1/PDL1-based inhibition of cytotoxic T-cell activity, among other processes. The team explored a wide set of parameters using global sensitivity analysis to see which factors most influence tumor behavior. As a validation step, sensitivity analysis recovered known correlates of better SCLC outcomes, including the importance of macrophage-mediated cytotoxic T-cell recruitment. Comparing TumorSim and GranSim highlighted shared drivers of lesion dynamics. Both models unexpectedly predicted a two-phase formation process, with an abrupt shift in lesion behavior when adaptive immune cells arrive from lung-draining lymph nodes. Simulations also showed a complex role for CCL5: later in tumor growth it associates with better control, but early on it can act pro-tumor by recruiting regulatory T cells. Finally, like virtual TB granulomas, TumorSim tumors tended to grow larger when immunosuppressive mechanisms outweighed pro-inflammatory responses.

These results matter because they point to common principles underlying very different lung diseases. By showing a similar two-phase pattern in both TB granulomas and SCLC tumors, the models suggest that timing of immune arrival and the balance between suppressive and inflammatory signals can steer lesion outcomes. The finding about CCL5 illustrates how the same molecule can help or hurt depending on when and where it acts, and it highlights regulatory T cells as a potential early influence on tumor fate. Because TumorSim is built on GranSim, researchers now have a linked computational framework to test hypotheses about lung lesion evolution that would be difficult or slow to probe in patients. The authors propose TumorSim as a foundation for future studies of lung tumor-immune dynamics, including work on immunotherapeutics and anti-cancer drugs, and as a tool to study heterogeneity in disease progression and treatment response.