Anticancer Peptide Prediction via Multi-Kernel CNN and Attention Model.

Area of Science:

• Biomedicine
• Artificial Intelligence
• Computational Biology

Background:

• Modern lifestyles increase cancer incidence.
• Anticancer peptides (ACPs) are crucial for cancer cell elimination and therapy.
• Research interest in ACPs is growing due to their therapeutic potential.

Purpose of the Study:

• To develop an effective computational tool for identifying anticancer peptides.
• To leverage deep learning for enhanced ACP prediction.
• To improve the interpretability and performance of ACP identification models.

Main Methods:

• Development of the ACP-MCAM model, integrating a multi-kernel convolutional neural network (CNN) and an attention mechanism.
• Automatic learning of adaptive embeddings and context sequence features for ACPs.
• Model visualization for enhanced interpretability.

Main Results:

• ACP-MCAM significantly outperformed existing state-of-the-art models in identifying anticancer peptides.
• Different encoding schemes were evaluated, showing varying impacts on model performance.
• Method parameter optimization was conducted to refine the model's efficacy.

Conclusions:

• The ACP-MCAM model, combining multi-kernel CNN and self-attention, demonstrates superior performance in anticancer peptide identification.
• This work offers novel research directions for anticancer peptide prediction.
• The study fosters interdisciplinary advancements in artificial intelligence and biomedicine.