CACPP: A Contrastive Learning-Based Siamese Network to Identify Anticancer Peptides Based on Sequence Only

Anticancerpeptides (ACPs) recently have been receiving increasingattention in cancer therapy due to their low consumption, few adverseside effects, and easy accessibility. However, it remains a greatchallenge to identify anticancer peptides via experimental approaches,requiring expensive and time-consuming experimental studies. In addition,traditional machine-learning-based methods are proposed for ACP predictionmainly depending on hand-crafted feature engineering, which normallyachieves low prediction performance. In this study, we propose CACPP(Contrastive ACP Predictor), a deep learning framework based on theconvolutional neural network (CNN) and contrastive learning for accuratelypredicting anticancer peptides. In particular, we introduce the TextCNNmodel to extract the high-latent features based on the peptide sequencesonly and exploit the contrastive learning module to learn more distinguishablefeature representations to make better predictions. Comparative resultson the benchmark data sets indicate that CACPP outperforms all thestate-of-the-art methods in the prediction of anticancer peptides.Moreover, to intuitively show that our model has good classificationability, we visualize the dimension reduction of the features fromour model and explore the relationship between ACP sequences and anticancerfunctions. Furthermore, we also discuss the influence of data setconstruction on model prediction and explore our model performanceon the data sets with verified negative samples.

Automated summary

Anticancer peptides (ACPs) have gained attention in cancer therapy for their low consumption, few side effects, and easy accessibility. However, identifying anticancer peptides through experimental approaches is challenging and time-consuming. Traditional machine-learning-based methods for ACP prediction have low performance. In this study, a deep learning framework called CACPP is proposed, based on a convolutional neural network and contrastive learning, to accurately predict anticancer peptides. The comparative results on benchmark datasets demonstrate that CACPP outperforms state-of-the-art methods in predicting anticancer peptides. Visualization of feature dimension reduction and exploration of the relationship between ACP sequences and anticancer functions are also discussed.