DeepACP: A Novel Computational Approach for Accurate Identification of Anticancer Peptides by Deep Learning Algorithm

Area of Science:

• Biochemistry
• Computational Biology
• Drug Discovery

Background:

• Cancer poses a significant threat to human health.
• Accurate prediction of anticancer peptides (ACPs) is crucial for developing new anticancer drugs.
• While deep learning models show promise, the optimal architecture for ACP classification remains unclear.

Purpose of the Study:

• To systematically compare convolutional, recurrent, and convolutional-recurrent neural networks for ACP prediction.
• To develop and validate a deep learning tool, DeepACP, for predicting ACP activity.
• To enhance understanding of deep learning models in peptide identification.

Main Methods:

• Systematic exploration of three deep learning architectures: convolutional neural networks (CNNs), recurrent neural networks (RNNs), and CNN-RNN hybrid networks.
• Implementation of a sequence-based deep learning tool, DeepACP, utilizing the best-performing architecture.
• Performance comparison against existing ACP prediction methods.

Main Results:

• Recurrent neural networks, specifically those with bidirectional long short-term memory cells, demonstrated superior performance in distinguishing ACPs from non-ACPs.
• The developed DeepACP tool accurately predicts the anticancer potential of peptides.
• DeepACP exhibited enhanced performance compared to several existing prediction methods.

Conclusions:

• The study identifies bidirectional long short-term memory recurrent neural networks as the optimal architecture for ACP prediction.
• DeepACP serves as an effective, sequence-based tool for identifying potential anticancer peptides.
• The findings support the application of deep learning in proteomics and accelerate the development of novel therapeutics.