Peptide Resources

KELM-CPPpred is a kernel extreme learning machine (KELM)-based prediction model for cell-penetrating peptides, integrating amino acid composition, dipeptide composition and other features.

A high-throughput anti-inflammatory peptide (AIPs) predictor based on parallel deep neural networks, balancing sequence data via SMOTETomek to address challenges in precise AIPs classification for traditional ML algorithms.

LAMP is a linked database for antimicrobial peptides (AMPs), designed to facilitate the discovery and design of new antimicrobial agents. The current version contains 5,547 entries, including 3,904 natural AMPs and 1,643 synthetic peptides. Integrating detailed antimicrobial activity and cytotoxicity data, it supports keyword and combinatorial condition searches, with cross-linking and top similar AMP recommendation functions to assist in analyzing AMP structure-activity relationships. This accelerates the development of new AMPs with high antimicrobial activity and low cytotoxicity, promoting translation from basic research to clinical/preclinical trials.

A sequence-based predictor for anti-angiogenic peptides (AAPs) identification, achieving high-precision prediction via machine learning models. The model transforms each peptide sequence into a 4335-dimensional numeric vector based on 58 feature types, employs a heuristic algorithm for feature selection, and optimizes hyperparameters of six machine learning models for the selected feature subset. 

A computational framework for generating anticancer peptides (ACPs) by integrating Wasserstein Autoencoder (WAE) generative model and Particle Swarm Optimization (PSO) forward search algorithm, guided by an attribute predictive model. Compared to VAE and GAN, WAE demonstrates lower perplexity and reconstruction loss during training, with semantic connections in its latent space accelerating PSO's controlled generation process.