ACPNet: A Deep Learning Network to Identify Anticancer Peptides by Hybrid Sequence Information.

Area of Science:

• Computational Biology
• Bioinformatics
• Drug Discovery

Background:

• Cancer poses a significant global health threat, with drug resistance and side effects complicating treatment.
• Anticancer peptides (ACPs) show promise as novel therapeutic agents due to their efficacy and reduced side effects.
• Experimental identification of ACPs is resource-intensive, necessitating efficient computational prediction methods.

Purpose of the Study:

• To develop and validate a novel deep learning model, ACPNet, for accurate prediction of anticancer peptides.
• To improve upon existing computational methods for distinguishing anticancer peptides from non-anticancer peptides.

Main Methods:

• ACPNet integrates peptide sequence information, physicochemical properties, and auto-encoding features.
• A hybrid deep learning architecture combining fully connected and recurrent neural networks is employed.
• The model was trained and evaluated on the ACPs82 dataset and an independent validation set.

Main Results:

• ACPNet demonstrated improved performance metrics, including 1.2% Accuracy, 2.0% F1-score, and 7.2% Recall compared to existing methods.
• The model achieved balanced performance, including a strong Matthews correlation coefficient.
• Validation on an independent dataset showed high accuracy, with only 1 out of 20 known ACPs misclassified.

Conclusions:

• ACPNet is an effective deep learning tool for accurately distinguishing anticancer peptides from non-ACPs.
• The model's performance suggests its utility in accelerating the identification and development of novel ACPs for cancer therapy.
• This computational approach can aid in overcoming the limitations of experimental screening for ACP discovery.