ACP-ESM2: Enhancing Anticancer Peptide Prediction With Pre-Trained Protein Language Models

Area of Science:

• Biochemistry
• Computational Biology
• Bioinformatics

Background:

• Anticancer peptides (ACPs) are gaining attention for targeted cancer therapy due to low toxicity.
• Traditional experimental identification of ACPs is laborious and time-consuming.
• Existing deep learning methods for ACP prediction require further optimization.

Purpose of the Study:

• To develop a highly accurate deep learning framework for predicting anticancer peptides.
• To leverage evolutionary information and convolutional neural networks for enhanced ACP identification.

Main Methods:

• Utilized the Evolutionary Scale Modeling 2 (ESM2) pre-trained model to capture protein sequence evolutionary information.
• Integrated ESM2 with a convolutional neural network (CNN) for pattern detection.
• Developed the ACP-ESM2 framework for anticancer peptide prediction.

Main Results:

• ACP-ESM2 demonstrated significant performance improvements over existing methods on the Test1 dataset (ACC, SN, SP, MCC).
• Achieved 97.6% accuracy on the Test2 dataset, indicating high robustness.
• Showcased superior efficiency and precision in predicting anticancer peptides.

Conclusions:

• ACP-ESM2 represents an efficient and precise computational tool for anticancer peptide prediction.
• The framework offers a valuable alternative to experimental methods, accelerating drug discovery.
• Highlights the potential of combining large pre-trained models with CNNs for biological sequence analysis.