Deciphering the virus-host dialogue: Progress and prospects of AI-powered single-cell and spatial multi-omics in analyzing the hepatitis B immune microenvironment

Hepatitis B virus (HBV) infection affects nearly 300 million people worldwide, and its long-term immune escape and chronic inflammation are closely associated with the development of liver cirrhosis and hepatocellular carcinoma (HCC). However, HBV-related immunopathology often manifests as the combined effects of continuous changes in cell states, cross-layer regulatory networks, and tissue spatial architecture, making it difficult for traditional bulk omics to unravel the key mechanisms. The development of single-cell omics and spatial omics has significantly improved the resolution of analysis, but characteristics such as high noise, high dimensionality, and missing modalities pose challenges for data integration and statistical inference. In recent years, artificial intelligence (AI) methods, represented by deep learning, have provided new toolkits for unified representation learning, multi-modal integration, trajectory/directional inference, and spatial structure identification, enabling researchers to derive more consistent and reusable immune representations across multiple scales, such as cell-epigenetic-spatial niche. Current studies suggest that AI-assisted analysis helps to characterize the continuous spectrum of HBV-specific CD8 + T cell exhaustion, the spatial coupling features of inflammation and fibrosis, and provides a basis for proposing testable targets and patient stratification clues. This review surveys recent progress in AI-enabled HBV multi-omics integration and considers future directions, including improved data harmonization, greater model interpretability, and the establishment of an iterative validation framework that connects computational modeling with spatial and functional verification. Such an approach may provide a practical methodological basis for advancing research toward HBV functional cure.