Site Editor
Gregory J. Riely, MD, PhD
Gregory J. Riely, MD, PhD
Posted: Wednesday, November 13, 2024
A novel deep-learning approach called HistoTME could be an effective way to help predict response to immune checkpoint inhibitor (ICI) therapy in patients with non–small cell lung cancer (NSCLC), according to research presented at the 2024 Society for Immunotherapy of Cancer (SITC) Annual Meeting (Abstract 77). Tamara Jamaspishvili, MD, PhD, of SUNY Upstate Medical University, Syracuse, New York, and colleagues said their work employs artificial intelligence to access tumor microenvironment–related molecular composition information that is already accessible on hematoxylin and eosin (H&E)-stained pathology slides.
This work “bring[s] us closer to personalized immuno-oncology,” the researchers declared. The HistoTME model learns to predict the gene-expression levels of 30 cell type–specific tumor microenvironment signatures from H&E whole-slide images.
HistoTME-predicted signatures allowed the team to derive tumor microenvironment status—inflamed vs desert—which improved prognostication of patients receiving first-line ICI treatment (P = .0012). Patients with PD-L1 expression ≥ 50% showed improved survival as well (P = .0059).
Further subgroup analysis showed that the H&E-inferred tumor microenvironment status was predictive of overall survival outcomes in patients whose disease was PD-L1–absent (< 1%, P = .08) and PD-L1–low (1%-49%, P = .009)—but not PD-L1– high (≥ 50%, P = .85). “Additionally, the signatures could be utilized to accurately predict response, achieving an area under the receiver operating characteristics of 0.75 for predicting responses following first-line ICI treatment,” wrote Dr. Jamaspishvili and co-investigators.
The team’s methods included training HistoTME using H&E images and matched bulk transcriptomics data from The Cancer Genome Atlas (TCGA)-NSCLC cohort (n = 865). It was validated using an external cohort of Clinical Proteomic Tumor Analysis Consortium NSCLC (n = 333). Tumor microenvironment signatures were then used to predict response to immunotherapy. The methods were tested retrospectively using survival outcomes of an external cohort of 292 patients with NSCLC treated with immunotherapy.
Disclosure: No disclosure information was provided.
JNCCN Spotlights
