Peptide Resources

QSPred-FL is a predictor for detecting quorum-sensing peptides in large-scale proteomic data, leveraging feature representation learning to enhance prediction performance and providing a user-friendly web service.

A 2D Convolutional Neural Network-based model for anticancer peptide prediction, preprocessing peptide sequences by integrating one-hot encoding and physicochemical properties to capture spatial patterns.

sAMP-PFPDeep is a deep learning model for predicting short antimicrobial peptides (≤30 residues), converting sequences into three-channel images with positional, frequency, and 12 physicochemical features, leveraging VGG-16 and RESNET-50. VGG-16 achieves 87.37% testing accuracy.

sAMPpred-GAT is the first AMP predictor based on predicted peptide structures, constructing graph models with structural, sequence, and evolutionary information, using Graph Attention Network (GAT) for feature learning and fully connected networks for classification.

SATPdb is a manually curated database of structurally annotated therapeutic peptides, integrated from 22 public peptide databases/datasets (including 9 in-house datasets), currently containing 19,192 experimentally validated unique sequences (2-50 amino acids in length) with natural/non-natural/modified residues. Peptides are systematically categorized into 10 functional classes (e.g., 1,099 anticancer, 10,585 antimicrobial, 1,642 drug delivery, 1,698 antihypertensive), with 3D structure annotation via PDB and state-of-the-art methods (I-TASSER, HHsearch, PEPstrMOD). It supports structure/sequence similarity search, functional browsing, moonlighting peptide identification, and customized structure-activity retrieval, facilitating peptide-based drug research.