Anti-cancer Peptide Recognition Based on Grouped Sequence and Spatial Dimension Integrated Networks

Area of Science:

• Computational Biology
• Bioinformatics
• Drug Discovery

Background:

• Anti-cancer peptide (ACP) sequence prediction is challenging due to sequence diversity.
• Existing methods struggle to effectively capture complex sequence features.

Purpose of the Study:

• To propose GRCI-Net, an integrated network algorithm for enhanced anti-cancer peptide sequence prediction.
• To address the difficulties posed by the diversification of characteristic ACP sequences.

Main Methods:

• Implemented feature fusion reduction using Principal Component Analysis (PCA) on binary structure and K-mer sparse matrix features.
• Constructed a bidirectional long- and short-term memory network incorporating traditional and dilated convolutions for spatial feature extraction.
• Integrated grouping sequence features and spatial dimensional features using dense network layers and a sigmoid function for prediction.

Main Results:

• GRCI-Net achieved high prediction accuracy on two public datasets: ACP740 (0.8230) and ACP240 (0.8750).
• The model effectively handles diverse anti-cancer peptide feature sequences and learns contextual information.

Conclusions:

• GRCI-Net offers a more suitable and accurate approach for anti-cancer peptide sequence prediction.
• The developed algorithm demonstrates significant potential in advancing anti-cancer drug discovery efforts.