The following is a summary of “Prediction of radiosensitivity in non-small cell lung cancer based on computed tomography and tumor genomics: a multiple real-world cohort study,” published in the April 2025 issue of the Respiratory Research by You et al.

Tumor heterogeneity in radiosensitivity remains a major challenge in optimizing treatment efficacy for patients with non-small cell lung cancer (NSCLC). This study introduces a radiogenomic approach that integrates imaging and transcriptomic data, combined with insights from single-cell sequencing, to establish a predictive model capable of identifying intratumoral variation in radiosensitivity. The proposed framework aims to improve radiotherapy precision, minimize unnecessary toxicity, and support clinical decision-making.

This retrospective analysis utilized CT imaging and transcriptomic data from 454 patients diagnosed with NSCLC across multiple real-world clinical cohorts, all of whom underwent imaging prior to receiving radiotherapy. A subset of patients (n = 154) constituted the training set, from which the primary tumor region was delineated and segmented to extract radiomic features. These features were used to develop a radiogenomic signature, termed LCDigital-RT, by integrating a transcriptome-based radiosensitivity index.

Two independent external validation cohorts were used to assess model performance: the JXCH cohort (n = 74) and the GDPH cohort (n = 160). Additionally, single-cell RNA sequencing data were incorporated to explore the underlying biological mechanisms driving differential radiosensitivity at the cellular level, enabling a more refined interpretation of imaging-genomic correlations.

An initial predictive model based solely on radiomic features—referred to as pre-LCDigital-RT, demonstrated moderate accuracy in distinguishing radiosensitive versus radioresistant tumor regions. The area under the receiver operating characteristic curve (ROC) for this model was 0.759 in the training cohort, 0.728 in the JXCH cohort, and 0.745 in the GDPH cohort.

The fully integrated LCDigital-RT model showed improved predictive performance, with an AUC of 0.837 in the training set. This enhancement was consistently validated in external cohorts, with AUCs of 0.789 and 0.791 in the JXCH and GDPH datasets, respectively. The model effectively stratified patients into radiosensitive and radioresistant groups, revealing significant differences in tumor morphology and imaging features between these subgroups.

Further analysis using single-cell transcriptomics revealed distinct cellular compositions and signaling pathways associated with each radiosensitivity profile. These findings support the biological validity of the LCDigital-RT signature and underscore its potential value in translational oncology.

The LCDigital-RT model offers a robust, non-invasive method to predict radiosensitivity in patients with NSCLC by integrating radiomic and genomic features. This tool enhances the precision of radiotherapy planning by identifying intratumoral regions with differing radiation response profiles and guiding clinicians in tailoring treatment intensity. The incorporation of single-cell data strengthens the interpretability of radiogenomic features, bridging imaging, molecular biology, and clinical application. Ultimately, this approach holds promise for improving radiotherapy outcomes, reducing treatment-related toxicity, and informing personalized care strategies in lung cancer management.

Source: respiratory-research.biomedcentral.com/articles/10.1186/s12931-025-03202-z