Digitizing and Engineering Human Biology with Multimodal Biointerfaces and AI

Abstract

Living systems are dynamic, spatially organized, and continuously evolving. Understanding and engineering these systems require technologies that can maintain stable access to biological activity, deliver precise interventions, interpret complex multimodal data, and guide experimental actions through feedback. However, existing approaches often address these capabilities separately: conventional measurements provide only transient or incomplete access to living tissues, perturbations are frequently global or short-lived, and biological analysis and experimentation remain fragmented, labor-intensive processes. This dissertation develops bioelectronic, material, and artificial intelligence technologies that progressively expand how researchers interface with living systems from chronic sensing and modulation to spatially programmable modulation, and ultimately to agentic interpretation and execution of biological workflows. Chapter 2 introduces cyborg tissue-bioelectronic hybrid systems that integrate tissue-like, stretchable mesh nanoelectronics into developing three-dimensional organoids. Through a two-dimensional-to-three-dimensional reconfiguration process during organogenesis, distributed sensors and stimulators become seamlessly embedded throughout the tissue, enabling longitudinal, tissue-wide electrophysiological access while accommodating growth and structural remodeling. This chapter establishes a general protocol for fabricating and integrating mesh nanoelectronics with cardiac, neural, and pancreatic organoids. It then applies this platform to human stem cell-derived pancreatic organoids, enabling months-long, single cell-resolved recordings of α- and β-cell electrical activity. By combining spike-sorted electrophysiology with in situ electro-sequencing, the platform connects electrical signatures with transcriptionally defined cell identities. Longitudinal measurements reveal cellular electrical states associated with improved hormone responsiveness, while metabolic entrainment uncovers coordinated circadian electrical and secretory rhythms. Embedded electrical stimulators further enhance glucose-dependent α- and β-cell responses, demonstrating the ability to both monitor and modulate functional maturation within intact tissues. Chapter 3 extends biological modulation from prefabricated electronic devices to materials assembled directly within genetically specified cells. This chapter introduces optogenetic polymerization of conductive polyaniline and insulating poly(3,3’-diaminobenzidine) using genetically encoded photosensitizers. Light-controlled polymerization enables electrically functional materials to be deposited with genetic, spatial, and temporal specificity while limiting cytotoxicity. By altering neuronal membrane capacitance, conductive and insulating polymers produce stepwise decreases or increases in intrinsic neuronal excitability, respectively. This approach establishes a programmable material interface for writing persistent functional changes into living cells with cellular and subcellular precision. Chapter 4 introduces the Spatial Transcriptomics AI Agent, or STAgent, to address the analytical complexity generated by increasingly high-dimensional and spatially resolved biological measurements. STAgent integrates multimodal large language models with dynamic code generation, visual reasoning, specialized computational tools, scientific literature retrieval, and automated report generation. Applied to spatial transcriptomic measurements of human stem cell-derived pancreatic grafts across in vivo maturation, STAgent autonomously executes multistep analyses and transforms workflows that traditionally require weeks of expert effort into minutes. The agent identifies the progressive organization of endocrine cells into islet-like architectures, strengthening interactions among endocrine populations, and expansion of supporting mesenchymal networks. It further interprets spatially resolved gene programs associated with ion transport, secretory function, cell-cell communication, and extracellular matrix interactions, connecting computational observations to biologically grounded mechanistic explanations. Chapter 5 extends agentic AI from computational interpretation into physical biological experimentation. This chapter introduces Agentic Lab, an agentic-physical AI system that combines a virtual principal-investigator agent with specialized subagents for knowledge retrieval, protocol design, multimodal analysis, explainable morphological phenotyping, and laboratory observation. Through an augmented-reality interface, the system perceives experimental activities, provides context-aware instructions, identifies procedural deviations, and records experimental traces as long-term memory. In human pancreatic organoid differentiation, Agentic Lab generates experimental protocols, assists physical execution, identifies morphological heterogeneity associated with culture conditions, and proposes targeted adjustments to improve differentiation outcomes. The platform thereby connects experimental design, physical action, data interpretation, and iterative refinement within a unified human-AI workflow. In summary, this dissertation establishes complementary technologies for reading biological states, writing precise interventions, reasoning over multimodal evidence, and guiding experimental action. By integrating cyborg tissue-bioelectronic systems, optogenetically programmable materials, and agentic AI, this work advances biological engineering from passive observation toward interactive and adaptive engagement with living systems. Together, these platforms provide foundations for intelligent biointerfaces and laboratories that can continuously sense, interpret, modulate, and refine biological processes across molecular, cellular, and tissue scales.