Peptide Resources

BIOPEP-UWM™ (formerly BIOPEP) is a professional database for bioactive peptides, focusing on food-derived peptides and diet-related chronic disease prevention research. It supports continuous updates with new features like D-amino acid peptide annotation, batch processing, amino acid sequence to SMILES code conversion, and provides protease-released peptide prediction and active fragment quantitative analysis, serving as a data resource for nutraceutical and functional food research.

A predictor for anti-angiogenic peptides (AAPs) based on generalized linear model (glmnet), integrating three types of amino acid composition descriptors (e.g., SP, LSL motifs) to construct multi-dimensional feature space, with feature selection to reduce dimensionality and filter noise.

Brainpeps is a specialized database for blood-brain barrier (BBB) peptide transport, systematically integrating scattered data on peptide BBB permeability from literature, including positive/negative transport results. It supports retrieval by transport mechanisms, experimental methods (in vitro cell models/in vivo animal assays), and structural parameters (molecular weight, hydrophobicity). The database classifies and visualizes BBB assessment methods to facilitate analysis of peptide structure-BBB permeability relationships, providing data support for brain-targeted peptide drug design.

CACPP is a deep learning framework based on convolutional neural network (CNN) and contrastive learning for accurate prediction of anticancer peptides (ACPs). It extracts high-latent features from peptide sequences via TextCNN and employs a contrastive learning module to learn more discriminative feature representations, eliminating dependence on handcrafted feature engineering. Benchmark experiments show CACPP outperforms all state-of-the-art methods, with feature dimensionality reduction visualization and dataset construction impact analysis provided.

An AdaBoost-based classification tool for antidiabetic peptides (ADPs), specialized in differentiating peptides targeting type 1 diabetes (T1DM) from those for type 2 diabetes (T2DM). The model employs Lasso penalized feature selection to identify critical sequence features (amino acid composition, hydrophobicity, charge distribution, etc.), and confirms AdaBoost's optimal performance after comparing with logistic regression, SVM, etc.