Peptide Resources

AAPred-CNN is the first deep convolutional neural network (CNN)-based predictor for anti-angiogenic peptides (AAPs), challenging the conventional belief that deep learning requires massive labeled data. It achieves efficient prediction even with scarce experimentally validated AAP sequences by extracting spatial feature patterns via multi-layer convolutional architecture, integrating residue position encoding and local contextual information.

ACP-2DCNN is a deep learning-based method for anticancer peptide prediction, extracting key sequence features via Dipeptide Deviation from Expected Mean (DDE) and using 2D Convolutional Neural Network (2D CNN) for training and prediction.

ACP-BC is a three-channel end-to-end model improving anticancer peptide (ACPs) prediction via data augmentation. The first channel uses bidirectional LSTM for raw sequence feature extraction, the second converts sequences to chemical formulas with SMILES notation and employs chemical BERT for deep abstract features, and the third manually selects four effective features including dipeptide composition.

ACP-CapsPred is a two-stage computational framework for precise identification of anticancer peptides (ACPs) and characterization of their functional activities across different cancer types. Integrating protein language models, peptide evolutionary information, and physicochemical properties, it constructs multi-dimensional feature profiles and employs capsule networks for prediction.

ACP-check is a predictive model for anticancer peptides based on bidirectional long short-term memory network (BiLSTM) and multi-feature fusion, extracting sequence time-dependent information and integrating binary profiles, dipeptide composition, k-spaced amino acid group pairs, etc.