Peptide Resources

Hemolytik is a manually curated database of experimentally determined hemolytic and non-hemolytic peptides, compiled from numerous published research articles and various databases (e.g., Antimicrobial Peptide Database, Collection of Anti-microbial Peptides, Dragon Antimicrobial Peptide Database, and Swiss-Prot). The current release contains approximately 3,000 entries, including ~2,000 unique peptides whose hemolytic activities were evaluated on erythrocytes isolated from 17 different sources.

A machine-learning pipeline for identifying potent antimicrobial peptides by mining hundreds of billions of sequences in the virtual library of 6-9 amino acid peptides. Following a coarse-to-fine design, it integrates trainable modules for empirical selection, classification, ranking, and regression to narrow down the search space, yielding hexapeptides with strong activity against multidrug-resistant pathogens and penicillin-comparable efficacy in mouse models.

HybAVPnet is a hybrid network-based predictor for antiviral peptides (AVPs), integrating neural networks and traditional machine learning to achieve precise identification. Adopting a stacking-like architecture, it combines LSTM for capturing long-term sequence dependencies, CNN for extracting local evolutionary features, and SVM for fusing multi-dimensional prediction labels and probabilities.

iACP-DRLF is a LightGBM-based tool for predicting anticancer peptides (ACPs), integrating two LSTM-derived deep neural network embedding techniques (Soft Symmetric Alignment and UniRep) to enhance ACP discrimination.

iACP-FSCM is an anticancer peptide (ACPs) predictor based on Flexible Scoring Card Method (FSCM), integrating local and global sequence propensity scores for accurate identification while addressing complexity and interpretability issues of existing models.