/The LR, RF, SVM and KNN algorithms were assessed in conjunction with 'sparse', 'dense', 'BLOMAP' and 'threemers' encoding methods. In general, each classifier exhibited varying levels of performance when coupled with the different encoding methods and no single encoding method outperformed the others, as evidenced by the findings presented in Figure 3. For example, the 'threemers' encoding exhibited exceptional effectiveness in the classification of cardiovascular, immunomodulatory, and antioxidant classes, likely due to the presence of specific tripeptides that augmented the classification power. On the other hand, the 'dense' and 'BLOMAP' encodings yielded the most favorable MCC values for neuropeptides and antihypertensives, respectively. Despite such a heterogeneity, the 'sparse' and 'threemers' encoding methods yielded the highest number of classifiers with MCC values exceeding 0.5, and consequently, they were selected as the encoding methods for Neural Networks (NN) to reduce the time for computational training. Analogous trends were observed for NN models, with highly variable performance associated with individual feature types depending on the considered functional class. Overall, the variability in the optimal encoding method across all classes underscores the necessity for a robust benchmark encompassing a wide range of encoding strategies to ensure the comprehensive capture of relevant features in BPs classification.