Graph Neural Networks for Cell-Cell Interaction Modeling

Heterogeneous Graph Neural Networks for Multi-Cell Type Interactions

This research investigates how heterogeneous GNNs can model complex interactions between diverse cell types with distinct molecular signatures and biological behaviors. The work produces novel architectural designs that capture differential communication patterns across multiple cell populations, advancing our understanding of tissue-level coordination mechanisms.

Temporal Graph Networks for Dynamic Cell Differentiation Pathways

This research explores temporal graph neural networks that capture time-evolving cell state transitions and differentiation cascades during development and reprogramming. The investigation yields mechanistic insights into how cell fate decisions propagate through intercellular signaling networks over developmental timescales.

Message Passing Mechanisms for Paracrine Signaling Network Inference

This research develops advanced message passing algorithms that decode paracrine signaling patterns from single-cell transcriptomics and spatial proteomics data. The work produces quantitative models of secreted factor communication that reveal previously undetected ligand-receptor interactions governing cell behavior.

Graph Attention Mechanisms for Cell Microenvironment Context Modeling

This research investigates attention-based graph mechanisms that identify which neighboring cells and environmental factors are most influential on individual cell phenotypes and functions. The research produces interpretable models that disambiguate key regulatory relationships within spatially-organized tissue architectures.

Spatial Graph Convolutions for 3D Tissue Architecture Reconstruction

This research develops spatial graph convolutional approaches that reconstruct three-dimensional tissue organization and cell positioning from multiplexed imaging and sequencing data. The advancement produces novel computational methods for understanding how 3D spatial geometry constrains cell communication and tissue function.

Graph Contrastive Learning for Cell-Cell Interaction Feature Discovery

This research applies self-supervised contrastive learning frameworks to graph representations of cell interaction networks without labeled interaction data. The work yields unsupervised discovery of emergent interaction patterns and functional cell communities that traditional supervised methods cannot identify.

Equivariant Graph Neural Networks for Biomolecular Symmetry Preservation

This research develops equivariant GNNs that preserve geometric and permutation symmetries inherent in cell-cell interaction systems and molecular structures. The investigation produces theoretically-grounded models that improve prediction accuracy by incorporating fundamental physical and biological conservation principles.

Hypergraph Neural Networks for Multi-Cellular Complex Formation Dynamics

This research extends GNNs to hypergraph structures to model higher-order interactions where more than two cells coordinate simultaneously in morphogenetic fields and tissue niches. The work reveals non-pairwise interaction rules that govern collective cell behaviors and emergent tissue properties.

Graph Signal Processing for Intercellular Communication Pattern Classification

This research combines graph signal processing theory with neural network architectures to decompose and classify communication signals propagating through cell-cell interaction networks. The advancement produces novel biomarkers for distinguishing healthy from pathological intercellular communication states.

Interpretable Graph Neural Networks for Mechanistic Cell Signaling Discovery

This research develops explainable GNN architectures that extract human-readable mechanistic rules governing cell-cell interactions from complex biological networks. The work produces validated causal insights into signaling pathways that can guide experimental validation and therapeutic intervention design.