mHPpred: Accurate identification of peptide hormones using multi-view feature learning - ScienceDirect

Fig. 1. Overall workflow methodology of mHPpred. The schematic workflow shows the incorporation of Chou's five-step rule involved in the development of mHPpred: (1) Construction of non-redundant benchmark and independent datasets; (2 & 3) Construction of 286 baseline models using 26 feature descriptors and 11 machine learning (ML) classifiers; (4) Evaluation of the top-performing baseline models and features using three different multi-view learning: meta-learning, feature fusion learning, and integrative framework. Following extensive validations, the optimal model, mHPpred, was created utilizing a meta-learning approach; and (5) Webserver development. Fig. 2. Performance comparison of baseline models for each feature descriptor. The evaluation performances of baseline models for each descriptor during training are shown as follows: (A) area under the receiver operating characteristic (ROC) curve (AUC), (B) Matthews' correlation coefficient (MCC), and (C) accuracy (ACC) performance me

Fig. 1. Overall workflow methodology of mHPpred. The schematic workflow shows the incorporation of Chou's five-step rule involved in the development of mHPpred: (1) Construction of non-redundant benchmark and independent datasets; (2 & 3) Construction of 286 baseline models using 26 feature descriptors and 11 machine learning (ML) classifiers; (4) Evaluation of the top-performing baseline models and features using three different multi-view learning: meta-learning, feature fusion learning, and integrative framework. Following extensive validations, the optimal model, mHPpred, was created utili