Peptide Resources

PREDAIP is an Extremely Randomized Tree (ERT)-based predictor for anti-inflammatory peptides (AIPs), integrating multi-sequence feature descriptors with emRMR-SFS hybrid feature selection. It first ranks features via four enhanced minimal redundancy maximal relevance (emRMR) algorithms, then uses Sequential Forward Selection (SFS) to build seven classifier wrapper models for optimal feature subset screening, feeding into ERT. Outperforming existing tools on independent datasets, PREDAIP validates the necessity of multi-correlation criteria in emRMR for feature optimization.

PSRQSP is an interpretable predictor for quorum sensing peptides, leveraging propensity score representation learning to extract amino acid and peptide propensities for meta-predictor construction.

PTPAMP is a prediction tool for plant-derived antimicrobial peptides, constructing models based on multiple peptide features such as amino acid composition, dipeptide composition, and physicochemical attributes. It can classify query peptide sequences into four functional activities: antimicrobial (AMP), antibacterial (ABP), antifungal (AFP), and antiviral (AVP).

A prediction tool for anticancer peptides (ACPs) based on 19-dimensional feature vectors, reducing model dimensionality by extracting key sequence features to address performance bottlenecks caused by high-dimensional features in machine learning.

A deep learning tool integrating sequence generative adversarial nets, BERT, and MLP, screening novel antimicrobial peptides (AMPs) via AlphaFold2 structure prediction and molecular dynamics simulation. The NMR-verified peptide A-222 inhibits gram-positive/negative bacteria, with analogs showing 4-8× activity enhancement.