Developing interpretable graph neural network frameworks for analyzing brain functional connectivity from fMRI data. Central to this line of work is the Gated Graph Transformer (GGT), which integrates prior spatial knowledge with random-walk diffusion strategies and gating mechanisms to capture complex structural and functional relationships between brain regions. These methods use attention-based multi-view node feature embeddings to identify significant functional brain network biomarkers while maintaining strong interpretability.

Representative publications:

• Interpretable Cognitive Ability Prediction: A Comprehensive Gated Graph Transformer Framework for Analyzing Functional Brain Networks — IEEE Trans. Medical Imaging, 2024
• Brain Functional Connectivity Analysis via Graphical Deep Learning — IEEE Trans. Biomedical Engineering, 2022
• Ensemble Manifold Regularized Multi-Modal Graph Convolutional Network for Cognitive Ability Prediction — IEEE Trans. Biomedical Engineering, 2021

Collaborations: TReNDS Center (GSU/Georgia Tech/Emory), DICoN Lab (Boys Town National Research Hospital), Mind Research Network

Designing frameworks that fuse complementary brain imaging modalities—functional MRI, diffusion tensor imaging, and structural MRI—to achieve a more comprehensive understanding of brain organization. Using masking strategies, cross-attention mechanisms, and graph neural networks to weight neural connections across modalities, these methods improve prediction accuracy and reveal how brain structure, function, and connectivity jointly relate to cognition and neuropsychiatric disorders.

Representative publications:

• An Interpretable Cross-Attentive Multi-Modal MRI Fusion Framework for Schizophrenia Identification — NeuroImage: Reports, 2026
• Integrated Brain Connectivity Analysis with fMRI, DTI, and sMRI Powered by Interpretable Graph Neural Networks — Medical Image Analysis, 2025
• Cooperative Multiplex GNN for High-Grade Glioma Survival Prediction from Preoperative Multi-Modal Radiomics-Based Brain Networks — IEEE Trans. Medical Imaging, 2026

Building computational frameworks that leverage deep learning and large language models for genomic and epigenomic analysis, with a focus on Alzheimer's disease. This includes transductive learning for polygenic risk score variant prioritization with GPT-powered interpretation, transformer architectures for cross-tissue DNA methylation-based disease detection, and curated databases for brain disease cell-cell communication networks.

Representative publications:

• BrainGeneBot: A Framework for Variant Prioritization and Generative Pretrained Transformer-Informed Interpretation Across Polygenic Risk Score Studies — Briefings in Bioinformatics, 2025
• BDCD: A Comprehensive Brain Disease Cell-Cell Communication Database — Database, 2026

Current focus at UTHealth Houston (BSML, PI: Dr. Zhongming Zhao):

• Multi-omics data integration for disease genomics
• Epigenomic biomarker discovery for neurodegeneration

Developing AI-driven solutions for clinical applications including automated digital pathology analysis. This includes deep learning systems for whole-slide image analysis, automated scoring systems (PD-L1 CPS), and histopathology image classification using semi-supervised learning approaches.

Representative publications:

• AI-based automated PD-L1 CPS scoring (Merck internship, 2022) — Linked to KEYTRUDA® anti-PD-1 therapy, achieving AUROC > 0.98
• Graph Temporal Ensembling Based Semi-Supervised Convolutional Neural Network with Noisy Labels for Histopathology Image Analysis — Medical Image Analysis, 2020