Xinyuan Bi et al. develop a hyperspectral surface-enhanced Raman spectroscopy (SERS) technique called SERSomes to profile the metabolic phenotypes of various biofluids, including cell culture medium and human serum. This method offers rapid, low-cost, and high-sensitivity profiling for fundamental studies and clinical applications. SERSomes improve diagnostic accuracy with deep learning models and facilitate the discovery of metabolite biomarkers. The technique is validated through experiments on cell culture medium and human serum, achieving an accuracy of 80.8% on an internal test set and 73% on an external validation set for prostate cancer diagnosis. The method also predicts tissue-level aggressiveness and identifies potential biomarkers, demonstrating its potential for cancer liquid biopsies and broader clinical applications.Xinyuan Bi et al. develop a hyperspectral surface-enhanced Raman spectroscopy (SERS) technique called SERSomes to profile the metabolic phenotypes of various biofluids, including cell culture medium and human serum. This method offers rapid, low-cost, and high-sensitivity profiling for fundamental studies and clinical applications. SERSomes improve diagnostic accuracy with deep learning models and facilitate the discovery of metabolite biomarkers. The technique is validated through experiments on cell culture medium and human serum, achieving an accuracy of 80.8% on an internal test set and 73% on an external validation set for prostate cancer diagnosis. The method also predicts tissue-level aggressiveness and identifies potential biomarkers, demonstrating its potential for cancer liquid biopsies and broader clinical applications.