his study investigates the potential of Raman spectroscopy for liquid biopsy in prostate cancer by comparing serum and urine as biofluids for classification. Principal Component Analysis (PCA) was applied to Raman spectra from both serum and urine samples to differentiate between healthy individuals and prostate cancer patients. In serum, the PCA plot showed partial separation, with some overlap between healthy controls and prostate cancer patients. The first principal component (PC1) explained 77.23 % of the variance, indicating that global spectral differences are key to distinguishing between the two groups. For urine, the separation was more distinct, with PC1 explaining 68.58 % and PC2 contributing 17.3 % of the variance. Specific spectral markers, such as a peak at 1000 cm−1, were associated with disease presence in urine, suggesting its potential as a more reliable biofluid for Raman-based diagnostics. Classification performance was assessed using three machine learning models: k-Nearest Neighbors (kNN), Random Forest (RF), and Support Vector Machine (SVM). Urine samples consistently outperformed serum in the kNN model, achieving higher accuracy (0.87 vs. 0.81) and sensitivity (0.91 vs. 0.79). The RF and SVM models showed balanced performance for both biofluids, with SVM performing best for serum (accuracy: 0.88, sensitivity: 0.94). Overall, the results indicate that urine provides clearer separation and more consistent performance, making it a more promising biofluid than serum for Raman-based prostate cancer liquid biopsy.