We explored six different machine-learning (ML) algorithms, namely, support vector machine (SVM), random forest (RF), gradient boosting (GB), extremely randomized tree (ERT), AdaBoost, and k-nearest neighbor based on 51 feature set derived from nine different feature encodings for the prediction of AHTP models. While ERT-based trained models performed consistently better than other algorithms regardless of various feature encodings, we treated them as base-line predictors, whose predicted probability of AHTPs were further used as input features separately for four different ML-algorithms (SVM, RF, GB and ERT) and developed their corresponding meta-predictors using a two-step feature selection protocol. Subsequently, the integration of four meta-predictors through an ensemble learning approach improved the balanced prediction performance and model robustness on the independent dataset. For a given peptide, mAHTPred predicts its calss and probability values.
