AI-Enhanced Detection of Dynamic Structural Changes in Inflammatory Protein Interfaces: A Case Study of CD11b/Mac-1 Interactions

This study presents a novel AI-enhanced approach for detecting dynamic structural changes in inflammatory protein interfaces, focusing on CD11b/Mac-1, a critical integrin involved in leukocyte adhesion and inflammatory signaling. We implemented a graph convolutional network (GCN) architecture that represents protein interfaces as nodes with physicochemical properties and geometric relationships as edges, enabling detection of conformational transitions with <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{9 2. 8 \%}$</tex> accuracy. The methodology integrates molecular dynamics simulations with deep learning, identifying five key dynamic regions within CD11b/Mac-1 interfaces, including the MIDAS domain and <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\boldsymbol{\alpha} 7$</tex> helix, which demonstrate significant conformational plasticity. Structure-function correlation analysis revealed strong associations between specific interface conformations and inflammatory outcomes, with MIDAS domain conformation strongly correlating with ligand binding affinity (<tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathrm{r}=0.87, \mathrm{p}&lt;0.001$</tex>) and <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\boldsymbol{\alpha} 7$</tex> helix extension with NF- <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\boldsymbol{\kappa} \mathbf{B}$</tex> activation (<tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{r}=\mathbf{0. 7 9}, \mathbf{p}&lt;\mathbf{0. 0 0 1}$</tex>). Our approach demonstrated substantial improvements over traditional methods in both accuracy and computational efficiency, requiring 95.4 % less processing time compared to conventional molecular dynamics approaches. The identified dynamic interface regions and their correlation with inflammatory functions provide rational targets for therapeutic development and establish a methodological framework applicable to diverse protein systems requiring dynamic interface analysis.