Artificial intelligence for mechanistic understanding of hepatitis B virus

Chronic hepatitis B virus (HBV) persists through a compact proteome, deep reliance on host pathways, and a nuclear covalently closed circular DNA (cccDNA) reservoir that current antivirals rarely extinguish. This Mini Review synthesizes advances from 2020–2025 in which artificial intelligence (AI) augments mechanistic understanding of HBV rather than serving only predictive ends. We summarize (i) AI-enabled structural modeling that clarifies polymerase priming and HBx architecture; (ii) physics-informed and multiscale inference that links sparse measurements to replication and cccDNA kinetics; (iii) sequence-based learners that expose non-random host-genome integration contexts and mutational constellations associated with immune tolerance or escape; (iv) network-aware analyses that prioritize host dependencies and connect CRISPR perturbations to virus–host modules governing cccDNA transcriptional control; and (v) AI-assisted antiviral discovery that couples virtual screening with mechanism-anchored interpretation (e.g., capsid assembly modulators). Across these domains, AI sharpens hypotheses by mapping viral mutations and host factors to discrete steps of the life cycle, quantitatively elevating high-leverage processes such as nucleocapsid recycling and cccDNA silencing, and guiding structure- or phenotype-guided intervention design. We also outline practical constraints—data sparsity, cross-cohort heterogeneity, and interpretability—and propose priorities that couple computation and experiment: mechanism-aware gray-box models, causal and spatially resolved analyses, calibrated uncertainty and benchmarking across genotypes, and active-learning loops that maximize information gained per experiment. Framed this way, AI emerges as a mechanism-aware partner to experimental virology, accelerating routes toward durable functional cure through eradication or stable transcriptional silencing of cccDNA.