Peptide Resources

PlantPepDB is the first manually curated database focusing on plant-derived peptides, containing 3,848 entries including 2,821 protein-level experimentally validated peptides, 458 transcript-level evidence peptides, 530 predicted peptides, and 39 homology-inferred peptides. Integrating physicochemical properties (isoelectric point, hydrophobicity) and tertiary structure information, it supports simple, advanced, physicochemical, and amino acid composition searches to facilitate therapeutic potential research. Notably, data curation reveals numerous peptides exhibit multiple bioactivities (e.g., antibacterial + antioxidant).

pLM4ACE is a predictor for antihypertensive peptide screening based on protein language model (ESM-2 embeddings), trained with machine learning methods like logistic regression to achieve optimal performance among 65 classifiers.

PPTPP (Physicochemical Property-based Therapeutic Peptide Predictor) is a Random Forest-based tool designed to address the limitation of existing methods in simultaneously enabling high-quality generic prediction and informative physicochemical properties (IPPs) identification for therapeutic peptides. The model innovatively constructs a physicochemical property-associated feature encoding system, integrating amino acid composition, hydrophobicity indices, charge distribution, and other multi-dimensional parameters to enable efficient screening of various therapeutic peptides (antiviral, antidiabetic, etc.).

PractiCPP is a deep learning framework tailored for extremely imbalanced datasets in cell-penetrating peptide prediction, integrating hard negative sampling and advanced feature extraction modules.

PreAIP is a computational predictor for anti-inflammatory peptides (AIPs) developed by integrating multiple complementary features, leveraging random forest classifiers to systematically combine primary sequence, evolutionary, and structural information.