ACP-DL: A Deep Learning Long Short-Term Memory Model to Predict Anticancer Peptides Using High Efficiency Feature Representation
Cancer is a well-known dreadful killer of human beings health, which has led to countless deaths and misery. Anticancer peptides open promising perspective for the cancer treatment and have various attractive advantages. Conventional hands-on methods are expensive and inefficient to find and identify novel anticancer peptides. There is an urgent need to develop novel efficient measures to predict novel anticancer peptides. In this study, we proposed a deep learning Long-Short Term Memory (LSTM) neural network model, named ACP-DL, to effectively predict novel anticancer peptides. The efficient features exploited from peptides sequences are fed to train LSTM model. More specifically, to fully exploit protein sequence information, we developed an efficient feature representation approach by integrating binary profile feature and k-mer sparse matrix of the reduced amino acid alphabet. Then we implemented a deep LSTM model to automatically learn how to identify anticancer peptides and non-anticancer peptides. To our knowledge, this is the first time that the deep LSTM model has been applied to predict anticancer peptides. It was demonstrated by rigorous cross-validations experiments that the proposed ACP-DL remarkably outperformed other comparison methods with 81.48% accuracy at the AUC of 0.894 on benchmark dataset ACP740 and with an accuracy of 85.42% at the specificity of 89.94% and the AUC of 0.906 on dataset ACP240, respectively. In addition, we also contributed two anticancer peptides benchmark datasets ACP740 and ACP240 in this work.
Yi H-C, You Z-H, Zhou X, Cheng L, Li X, Jiang T-H, et al. ACP-DL: A Deep Learning Long Short-Term Memory Model to Predict Anticancer Peptides Using High-Efficiency Feature Representation. Molecular Therapy - Nucleic Acids. 2019;17:1-9. doi: 10.1016/j.omtn.2019.04.025.