Peptide Resources

ACP-ESM2 is a pre-trained classifier-based model for anticancer peptide (ACPs) prediction, composed of ESM2 pre-trained model, bidirectional GRU recurrent neural network, and fully connected layer. It feeds ACP sequences into ESM2 for dimension expansion, processes through bidirectional GRU, and generates outputs via fully connected layer.

ACP-ML is a machine-learning model for predicting anticancer peptides (ACPs), using DPC, PseAAC, CTDC, CTDT, and CS-Pse-PSSM features, followed by two-step feature selection with MRMD and RFE algorithms.

ACP-MS is an efficient model for predicting anticancer peptides (ACPs). It first uses the monoMonoKGap method to extract sequence features and convert them into digital features, then uses the AdaBoost model to select the most discriminative features, and finally introduces a stochastic gradient descent algorithm to identify anticancer peptide sequences.

ACP-OPE is a predictive tool for anticancer peptides (ACPs) integrating deep learning and machine learning, using ordinal positional encoding combined with handcrafted features for sequence representation. It employs a dual-channel deep learning module (BiLSTM + CNN) for temporal/spatial feature extraction, couples with LightGBM, and fuses results via voting.

ACPPred is a webserver for predicting anticancer peptides using SVM and protein relatedness measure features, assessing apoptotic domains and scanning for ACPs in proteins. Validated on multiple datasets, it achieves 96% accuracy and outperforms existing methods. It offers three modes (protein scan, multi-peptide, peptide mutation design) and is freely accessible via PERL CGI.