.png){kind=icon}
.png){kind=icon}
Return
DMVHP-IBS: Dynamic feature-integrated multi-modal prediction of virus-host protein interactions and the binding sites
.png){kind=icon}
PRE
AI
L
S
G
Y
Z
DOI：10.1016/j.artmed.2026.103423.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
• For protein structure data, we extract various chemical properties and dynamic information, including low-frequency vibrational modes and Dynamic Cross-Correlation Maps (DCCM), to construct the protein graph. • We propose a novel graph neural network architecture, JK-GCN, to extract structural information from viral and host proteins. Simultaneously, we utilize a pre-trained protein language model to process protein sequence data, thereby supplementing the deficiencies in structural information. • The GEICA method calculates the contribution weights of each amino acid to human-virus protein-protein interactions by using the VH-PPI scores and the gradient magnitudes of each amino acid node after each layer of graph convolution. Additionally, it leverages the attention mechanism of the pre-trained protein language model to further enhance our predictive results.
Keywords:
protein graph
graph neural network
protein-protein interactions
dynamic features
attention mechanism
Journal
IF:
6.2
Papers:
2.5K
Citations:
7.8K
Organization
No organization information available
Cited Papers