S196
Clinical - Biomarkers of clinical response
ESTRO 2026
radiotherapy (proPSMA): a prospective, randomised, multi-centre study. Lancet. 2020;395(10231):1208-1216 Keywords: prostate cancer, PSMA PET/CT, WB-MRI
Digital Poster 4011
non-invasive detection of prostate cancer via urinary carbohydrate metabolite profiling using mass spectrometry and machine learning Jeffrey Yungchuan Chao 1 , I-chuang Lu 2 1 radiation oncology, Taichung Veterans General Hospital, Taichung, Taiwan. 2 Chemistry, National Chung Hsing University, Taichung, Taiwan Purpose/Objective: Cancer cells exhibit distinctive metabolic alterations, particularly in carbohydrate metabolism, which is crucial for energy production and biosynthesis during proliferation. Conventional diagnostic methods for cancer are often time-consuming, costly, or insufficiently accurate, while some invasive procedures carry risks of complications. This study explored the feasibility of matrix-assisted laser desorption/ionisation mass spectrometry (MALDI-MS) combined with machine learning for the rapid and non-invasive detection of prostate cancer through analysis of urinary carbohydrate metabolites. We employed the rapid freeze-drying droplet (RFDD) technique, designed to improve reproducibility and signal intensity of polysaccharides, thereby overcoming key limitations of traditional MALDI-MS analysis. This approach aims to facilitate the identification of cancer-related metabolic changes and enhance the efficiency of clinical diagnosis and screening. Material/Methods: Fifty participants were recruited, and all provided written informed consent approved by the institutional review board. Voluntarily voided urine specimens were collected from three cohorts: healthy controls (no diagnosed cancer), patients with localised prostate cancer, and patients with confirmed distant metastatic prostate cancer. Two sample preparation methods were compared: the conventional dried droplet (DD) method and the RFDD method, both using 2,5-dihydroxybenzoic acid (2,5-DHB) as the MALDI matrix. In the RFDD method, 1 μ L of processed urine was deposited onto a stainless-steel plate, rapidly frozen in liquid nitrogen, and vacuum-dried. The dried samples were subsequently analysed by MALDI-MS at room temperature.
Results: This study evaluated the diagnostic potential of MALDI-MS by profiling urinary polysaccharides and metabolites. The RFDD method significantly enhanced carbohydrate metabolite detection, generating clearer and more distinct ion signal profiles. We were unable to analyse the first 11 urine specimens owing to the degradation of polysaccharides that occurred during the prolonged interval between collection and sample preparation. Mass spectrometry data from 39 patients were randomly divided into training (70%) and validation (30%) sets for machine learning model development. Models trained on RFDD-prepared samples achieved 77% accuracy in classifying prostate cancer with or without metastasis, compared with 40% accuracy for models trained on DD-prepared samples. The superior performance of the RFDD method likely results from improved detection of polysaccharides associated with altered carbohydrate metabolism in cancer cells.
Conclusion: MALDI-MS combined with machine learning provides a rapid, reliable, and non-invasive method for detecting
Made with FlippingBook - Share PDF online