0
0This research introduces a graph neural network (GNN) framework designed to analyze the complex spatial relationships within the tumor microenvironment of non-small cell lung cancer. By modeling localized cellular neighborhoods, the study successfully predicts patient survival with high accuracy, outperforming traditional density-based metrics. The analysis reveals that the prognostic impact of immune cells, such as CD8+ T cells, is heavily dependent on their spatial context and proximity to tumor or immunosuppressive cells. Using in-silico manipulations, the authors demonstrate how specific cell-to-cell interactions, like direct contact between CD8+ cells and tumor cells, correlate with better clinical outcomes. Ultimately, this approach identifies distinct TME states that balance immune activation and evasion, offering a new method for developing precision biomarkers.
References:
Hoebel KV, Lindsay JR, Altreuter J, Alessi JV, Weirather JL, Dryg I, Giobbie-Hurder A, Li Z, Yu KH, Awad MM, Rodig SJ, Lotter W. Graph neural network modeling of spatial tumor-immune interactions identifies prognostic cellular niches in non‑small cell lung cancer. NPJ Precis Oncol. 2026 Mar 7;10(1):158. doi: 10.1038/s41698-026-01314-3. PMID: 41794974; PMCID: PMC13096322.
