S2574
Radiobiology – Normal tissue radiobiology
ESTRO 2026
(2020) 1–11.. Keywords: Transit amplifying cell, alopecia, shh signaling
Digital Poster 3108
Integrating Multi-Omics and Machine Learning to Identify Key Biomarkers and Develop a Predictive Model for Radiodermatitis juan Deng, Chunmei Xia, Pei Wang Department of Oncology, Chongqing Hospital of Traditional Chinese Medicine, chongqing, China
Purpose/Objective: Radiodermatitis is one of the most common
complications of radiotherapy for cancer. Its typical symptoms include erythema, dry desquamation, and even moist desquamation and ulceration. These symptoms not only severely impact patients' quality of life but also frequently lead to interruptions in radiotherapy, thereby reducing the local tumor control rate. This study aims to utilize bioinformatics and machine learning methods to identify key biomarkers for the early diagnosis and risk prediction of radiodermatitis. Material/Methods: Gene expression datasets related to radiodermatitis (GSE154559, GSE300340) were downloaded from the NCBI GEO database. First, the Limma package was used to identify differentially expressed genes. Subsequently, Weighted Gene Co-expression Network Analysis was employed to identify key module genes most associated with the disease phenotype. To construct a highly robust predictive model, this study integrated a comprehensive machine learning framework. The preprocessed data was split, and 11 different algorithms covering various learning paradigms—including Ridge regression, Support Vector Machine, and XGBoost—were used for model building. Hyperparameters were optimized via cross- validation. Finally, a stacked integration strategy was applied to combine the optimal models, and core diagnostic biomarkers were selected based on feature importance rankings. Results: By integrating differential expression analysis and WGCNA, we successfully identified nine core genes closely associated with the development of radiodermatitis: CRISPLD2, GRIA3, GRIK2, ITGB8, NFATC2, PLCE1, RSRP1, SYTL2, and TNFRSF19. The integrated machine learning model built using these genes demonstrated excellent predictive performance. In the internal validation set, the model achieved an Area Under the Receiver Operating Characteristic Curve of 0.95, with both Accuracy and F1-score exceeding 0.90, indicating high reliability and stability
Conclusion: Transient, peri-irradiation Hh pathway inhibition modulates phase-specific radiosensitivity in vivo, lessens normal-tissue injury, and mitigates radiotherapy-induced alopecia by protecting TACs and facilitating TAC-driven repair. Targeting an organ- intrinsic proliferative signal provides a tissue-selective, mechanism-guided prophylaxis for regenerative epithelia and may be generalizable to other organs governed by local progenitor-cycling cues. References: 1.Huang W-Y, Lai S-F, Chiu H-Y, et al. Mobilizing transit- amplifying cell-derived ectopic progenitors prevents hair loss from chemotherapy or radiation therapy. 2017;77(22):6083-6096.2. Wei L, Leibowitz BJ, Wang X, et al. Inhibition of CDK4/6 protects against radiation- induced intestinal injury in mice. The Journal of clinical investigation. 2016;126(11):4076.3.Y.-C. Hsu, L. Li, E. Fuchs, Transit-amplifying cells orchestrate stem cell activity and tissue regeneration, Cell 157 (4) (2014) 935–949.4.C.-L. Chen, W.-Y. Huang, E.H.C. Wang, K.-Y. Tai, S.-J. Lin, Functional complexity of hair follicle stem cell niche and therapeutic targeting of niche dysfunction for hair regeneration, J. Biomed. Sci. 27 (1)
Made with FlippingBook - Share PDF online