Research led by Katharina Hoebel, MD, PhD, a research fellow at Harvard Medical School’s Dana-Farber Cancer Institute, has made significant strides in understanding non–small cell lung cancer (NSCLC) prognosis through spatial interactions analysis.
The study, presented at the 2024 AACR Annual Meeting, aimed to develop a predictive model that correlates expression patterns with NSCLC prognosis by utilizing biopsy samples from NSCLC patients. These samples were stained using a targeted multiplex immunofluorescence assay, and data from the samples were used to create a deep-learning model that incorporates cellular neighborhood graphs.
According to Hoebel, this novel model outperformed existing baseline measures, such as the PD-L1 tumor proportion score (TPS) and immune marker density-based models, by considering spatial interactions. The use of interpretability methods revealed three distinct clusters enriched in PD-L1, with one cluster showing enrichment in PD-L1–positive tumor cells and two clusters enriched in PD-L1–positive immune cells.
Notably, the study found that clusters enriched in PD-L1–positive immune cells correlated with favorable survival outcomes. Specifically, when PD-L1–positive immune cells were dominant in a cluster or grouped with other immune cells like CD8- and PD-1–positive cells, superior survival outcomes were observed.
This groundbreaking research highlights the importance of considering spatial interactions in developing predictive models for NSCLC prognosis and could potentially lead to more accurate and personalized treatment strategies in the future.
