MRG-H: A hierarchical framework for autonomous tissue regeneration via transcriptome-conditioned morphogenetic learning
July 13
Work on autonomous tissue repair has largely concentrated on single-axis pathway activation within highly controlled organoid systems. Yet, clinically relevant regeneration requires coordinated multicellular remodeling over protracted developmental windows and robust adaptation to the intrinsic heterogeneity of patient-derived tissues. These demands remain difficult to meet with existing rule-based protocols or conventional end-to-end stimulation regimes.
To bridge this gap, we introduce a hierarchical framework capable of orchestrating complex, long-horizon regenerative programs. The architecture employs a high-level controller that plans directly in transcriptomic space, producing gene-level or corrective morphogen instructions that steer the repair process. A low-level controller then translates each instruction into spatiotemporal patterns of paracrine, mechanical and bioelectric cues, thereby driving cell-scale trajectories toward the desired macro-architecture. The high-level layer continuously monitors emergent tissue states through live spatial RNA profiling, issuing adaptive commands that compensate for deviations introduced by the low-level layer or by stochastic microenvironmental forces.
We validated the framework in ex vivo digit explants, a widely used model of appendage regeneration. Across eight independent murine samples, the system achieved complete structural and functional restoration without manual intervention. Ablation studies confirmed that removal of either hierarchical tier markedly reduced rescue rates and left residual malformations, underscoring the necessity of both transcriptomic planning and localized actuation. Notably, the hierarchical strategy enabled rapid recovery from vascular occlusion or mispatterned ossification events that routinely derail conventional, monolithic stimulation protocols.
These findings establish step-level autonomy in a complex regenerative context, marking a substantive advance toward programmable, patient-specific tissue repair systems suitable for translational deployment.