MultiRepPI: a cross-modal feature fusion-based multiple characterization framework for plant peptide-protein...

Area of Science:

• Plant biology
• Computational biology
• Bioinformatics

Background:

• Plant peptide-protein interactions (PepPI) are vital for growth, development, immunity, and adaptation.
• Existing computational methods struggle with integrating multimodal data (sequence, structure, disorder) and capturing cross-dependent features for accurate PepPI prediction.

Purpose of the Study:

• To develop an efficient computational framework, MultiRepPI, for predicting plant PepPI by integrating multimodal information and cross-modal feature fusion.
• To improve the characterization of peptide and protein features for better prediction of their interactions.

Main Methods:

• Developed a multiple characterization framework (MultiRepPI) incorporating cross-modal encoding (CME), cross-modal attention (CMA), and disordered feature extraction (DFE) modules.
• CME fuses CNNs, RNNs, and feature enhancement for multi-scale feature extraction.
• CMA uses bi-directional attention and gating mechanisms to mine interaction patterns and binding sites.
• DFE extracts dynamic features from disordered protein regions.

Main Results:

• MultiRepPI demonstrated significant improvements in prediction performance and binding residue recognition compared to state-of-the-art methods on a benchmark dataset.
• The framework effectively integrates sequence, structure, and disorder properties for enhanced PepPI prediction.

Conclusions:

• MultiRepPI offers a reliable tool for efficient plant PepPI prediction, advancing plant biology research.
• The framework provides a foundation for future plant biology studies and peptide drug development.