PDeepPP is an integrated deep learning framework that enables robust identification of bioactive peptides (BPs) and protein post-translational modifications (PTMs) across a wide range of peptide features. It is designed by integrating existing pre-trained protein language models with a hybrid transformer-convolution architecture, and systematically extracts global and local sequence features by leveraging comprehensive benchmark datasets and implementing strategies to address data imbalance. Through extensive analysis including dimensionality reduction and comparative studies, PDeepPP demonstrates robust and interpretable peptide representations, achieving state-of-the-art performance on 25 out of 33 biological identification tasks. In particular, it achieves high accuracy in antimicrobial (0.9726) and phosphorylation site (0.9984) identification, 99.5% specificity in glycosylation site prediction, and a significant reduction in false negatives in antimalarial tasks. PDeepPP enables large-scale accurate peptide analysis to support biomedical research and discovery of novel therapeutic targets for disease treatment. All codes, datasets, and pre-trained models are publicly available via GitHub and Hugging Face.
