Proteotranscriptomic Biomarker Identification for Radiation Response Prediction In Non-Small Cell Lung Cancer Cell Lines
Abstract
Purpose
Predicting tumor radiosensitivity remains a major challenge in precision radiotherapy due to incomplete concordance between transcriptomic alterations and functional protein expression. This study aims to develop an integrated transcriptomic–proteomic framework to identify concurrent biomarkers predictive of radiation response, quantified by survival fraction at 2 Gy (SF2), in non-small cell lung cancer (NSCLC).
Methods
RNA sequencing and data-independent acquisition mass spectrometry (DIA-MS) proteomic data were collected from 73 and 46 NSCLC cell lines, respectively. After preprocessing, 1,605 shared genes were retained for analysis. Feature selection was performed using Lasso regression with frequency-based ranking under 5-fold cross-validation repeated ten times. Support vector regression (SVR) models were constructed using transcriptome-only, proteome-only, and combined feature sets. Model performance was evaluated using coefficient of determination (R2) and root mean square error (RMSE). Correlation analyses assessed RNA-protein concordance and relationships between selected biomarkers and SF2.
Results
RNA and protein expression levels demonstrated significant but moderate concordance (median Pearson’s r = 0.363), highlighting the complementary information captured by each modality. Independent feature-selection pipelines identified 20 prioritized gene signatures per modality, with concurrent biomarkers including KDM2A, PSIP1, and PTBP2. Models trained on single-omic features showed limited cross-omic generalizability, whereas the combined transcriptomic-proteomic model achieved balanced predictive performance (R2 = 0.461, RMSE = 0.120 for transcriptomic prediction; R2 = 0.604, RMSE = 0.111 for proteomic prediction). Several identified biomarkers, including KDM2A, GOT2, and CCDC9, exhibited negative correlations with SF2, suggesting associations with radiosensitizing effects.
Conclusion
This study presents a proteotranscriptomic modeling framework for predicting radiation response in NSCLC. By integrating transcriptomic and proteomic data, the proposed approach improves predictive robustness and biological interpretability compared with single-omic models. The identified concurrent biomarkers capture both regulatory and functional determinants of radiosensitivity, supporting their potential translational value for precision radiotherapy.