Peptide Resources

AIPs-DeepEnC-GA is a deep ensemble computational model for precise prediction of anti-inflammatory peptides (AIPs), achieving efficient prediction through innovative feature encoding and genetic algorithm optimization. The model employs n-spaced dipeptide position-specific scoring matrix (NsDP-PSSM) and pseudo position-specific scoring matrix (PsePSSM) to embed evolutionary features, combined with reduced-amino acid alphabet (RAAA-11) and composite physiochemical properties (CPP) to construct multi-dimensional feature space, using minimum redundancy maximum relevance (mRMR) for optimal feature subset selection.

AIPs-SnTCN is an anti-inflammatory peptide predictor based on an improved Self-Normalized Temporal Convolutional Network (SnTCN), integrating Skip-gram word embedding, BERT encoding, and Conjoint Triad Features (CTF) for multi-dimensional feature representation, optimized by SVM-RFE feature selection.

AIPStack is a two-layer stacking ensemble model for predicting anti-inflammatory peptides (AIPs). The model characterizes peptide sequences by hybrid features fused from two amino acid composition descriptors, and constructs a stacking ensemble model with random forest and extremely randomized tree as base classifiers and logistic regression as the meta-classifier.

AMAP is a machine learning model designed for predicting peptide biological activities, focusing on antimicrobial activity, using multi-label classification to identify 14 functional types (antimicrobial, antiviral, anticancer, etc.). Compared with traditional methods, it improves prediction accuracy through sequence feature integration and classification strategy optimization, validated by 10-fold cross-validation, benchmarking on recent experimentally verified peptides (not in training set), and performance comparison with existing AMP predictors.

AMP-BERT is an antimicrobial peptide classification model based on fine-tuned BERT architecture, extracting structural/functional features from peptide sequences to identify AMPs. It uses attention mechanisms for interpretable feature analysis, outperforms existing methods on external datasets, with public code and datasets.