ACP-MHCNN: an accurate multi-headed deep-convolutional neural network to predict anticancer peptides

Area of Science:

• Computational biology
• Bioinformatics
• Cancer research

Background:

• Chemotherapy for deadly cancers has limitations including low specificity and significant side effects.
• Anticancer peptides (ACPs) show promise as targeted cancer therapeutics with fewer adverse effects.
• Experimental identification of ACPs is costly and time-consuming, necessitating efficient computational methods.

Purpose of the Study:

• To develop a novel computational model for accurate Anticancer Peptide (ACP) identification.
• To extract and integrate discriminative features from diverse data sources efficiently.
• To provide a viable, cost-effective alternative to experimental ACP discovery.

Main Methods:

• Proposed a multi-headed deep convolutional neural network (ACP-MHCNN) model.
• Extracted sequence, physicochemical, and evolutionary features using numerical peptide representations.
• Employed cross-validation and independent dataset testing for rigorous evaluation.

Main Results:

• ACP-MHCNN demonstrated superior performance in ACP identification compared to existing models.
• Achieved significant improvements in accuracy (6.3%), sensitivity (8.6%), specificity (3.7%), precision (4.0%), and MCC (0.20).
• The model effectively integrates multi-source peptide information while managing parameter overhead.

Conclusions:

• ACP-MHCNN offers a highly effective computational tool for identifying anticancer peptides.
• The model's performance surpasses state-of-the-art methods, paving the way for accelerated ACP drug discovery.
• Open-source code and an online predictor are available for broader research application.