Zhenan Bao

Modelling human biology is difficult precisely because many of the responses that matter most emerge not within a single tissue or cell type, but at the boundaries where matrix, immune and systemic signals converge in ways that are hard to recreate in isolation. Conventional in vitro systems and animal studies have advanced our understanding of many physiological processes, but they still consistently fall short at these interfaces. In this Review, we discuss how engineered materials, organoids, microphysiological systems and computational models can be used together to build more informative human model systems. We first consider engineered materials as tools for controlling the local environment around cells, including matrix composition, stiffness, geometry, transport and time-dependent cues. We then examine organoids, assembloids and microphysiological system platforms, with emphasis on functional benchmarking, vascularization, immune integration, maturation, reproducibility and biological variation across donors and conditions. Finally, we discuss computational models, including spatial atlases, generative models and digital twins, as tools for organizing measurements, testing mechanisms and improving prediction. A key insight is that no current platform fully closes the loop between sensing, tissue modelling, computational calibration and feedback-guided intervention. The central challenge now is to build models whose complexity is justified by the human question they are designed to answer.

Advances in scientific AI have accelerated hypothesis generation, prediction, and experimental design, but translating these capabilities into reliable physical experimentation remains a major bottleneck in scientific discovery. Although automation, high-throughput experimentation, foundation models, and cloud infrastructure have accelerated individual stages of the scientific workflow, they have not unified the discovery process. Scientists must still coordinate between computational models and physical laboratory systems to translate scientific designs into executable experiments and ensure their results are reliable. We argue that the next generation of laboratories will be agentic: environments in which scientists, AI systems, equipment and robotic platforms operate as collaborative discovery partners. We propose an architecture built around two complementary components: an agentic foundation that translates scientific designs into coordinated execution plans, and an evolving laboratory world model that represents physical state, predicts action outcomes, and incorporates experimental feedback. Together, they connect computational reasoning to observable and verifiable physical execution. Multimodal perception, robotics, and human–AI interfaces can make laboratory actions and evolving experimental conditions accessible to these systems, supporting adaptive execution and learning. We further propose an evaluation framework centered on physical grounding, execution fidelity, scientific validity, and adaptive coordination under human oversight, rather than autonomy alone. This Perspective outlines how agentic laboratories could connect scientific reasoning, physical experimentation, and trustworthy evidence within a human-in-the-lead discovery loop.

The viscoelastic response of phase-separated polymers systems may deviate from bulk values at the nanoscale. However, current measurement techniques lack the resolution to unmask local variations. We previously demonstrated the advantages of using a focused, pulsed infrared (IR) laser to photothermally heat a sample locally during nanomechanical measurements with an atomic force microscope (AFM), with heating controlled by varying the laser repetition rate. Herein, we demonstrate this IR heating coupled with nano-Dynamic Mechanical Analysis (nDMA). First, we measured the storage and loss modulus of a homogenous copolymer at varying IR repetition rates (i.e., at varying sample temperature), demonstrating a viscoelastic modulus response that is typical near the glass transition temperature. Further, we varied the IR laser repetition rate while modulating the AFM probe between 5 Hz and 100 Hz. With time-temperature superposition, modulus master curves were constructed through the Williams-Landel-Ferry relation, and the shift factors of IR heating-coupled nDMA curves were used to estimate the relation between IR repetition rate and surface temperatures. Lastly, this technique was applied to a phase-separated polymer blend to demonstrate the utility of the technique to decouple the viscoelastic behavior of different phases by leveraging the high spatial resolution of AFM.

Li metal is a promising anode for next-generation batteries, but its high per-cycle capacity loss hinders its commercial adoption. A major contributor is the corrosion of Li during calendar aging, which accelerates the generation of degradation byproducts, such as the solid electrolyte interphase (SEI) and isolated Li (i-Li). Here, we reveal that Li corrosion is spatially heterogeneous: five times more capacity is lost at the edge overhang area of the current collector than at the center in the coin cell form factor. To address this, we developed a selected-area insulating layer (SAIL) that prevents Li plating in the overhang, yielding a 94% reduction in corrosion-induced capacity loss. Electron microscopy and titration gas chromatography were employed to visualize the morphological heterogeneity and quantify the disconnected Li loss at the center versus the overhang. Preventing dendritic Li plating in the overhang of the current collector can therefore substantially mitigate calendar aging-induced capacity loss.

Fluorinated ether electrolytes are promising for high-energy-density lithium metal batteries, yet their translation to larger-scale production requires systematic control of solvent impurities generated during synthesis and purification. Using 2-(2-(2,2-difluoroethoxy)ethoxy)–1,1,1-trifluoroethane (F5DEE) as a high-performance model solvent, we compare gram- and kilogram-scale batches to correlate impurity origin, purification, and electrochemistry. Although upscaled F5DEE shows >99% purity by standard techniques, ppm-level moisture and organic impurities reduce Li Coulombic efficiency and increase cell variability. We classify impurities by generation pathway into retained intermediates, starting-material-derived analogues, side-reaction products, and purification-introduced residues. Distillation, adsorption, and reactive purification implicate water, alkoxide-derived byproducts, and amines as key performance-limiting species. However, excessive purification generates secondary impurities that degrade performance, revealing a removal–generation trade-off. Thermal, electrochemical, operando microcalorimetry, and online mass spectrometry analyses reveal impurity-dependent current-collector corrosion, heat release, gas evolution, and full-cell degradation. This route-resolved, source-to-consequence workflow provides a transferable qualification framework for electrolyte solvent scale-up.

Helicenes attract attention due to their spring-shaped structures and conjugated backbones. Synthesizing longer helicenes allows for structure-property relationship investigation as additional coil layers are achieved. Therefore, developing helicenes as building blocks for oligomers and polymers will increase understanding of property tuning upon chain extension. We report the synthesis of discrete oligomers using the helicene N[5]HDI from lengths n=1 to n=5. Selective bromination of the oligomers allows for extension to longer oligomers, as well as the potential to use them as monomer units in polymerizations. The oligomers exhibit an odd-even effect, where odd oligomers prefer a ‘coiled’ conformation, while even oligomers prefer a ‘snake’ conformation. This results in improved orbital overlap and effective conjugation in odd oligomers, and small changes in electronics but increased aggregation in even oligomers. The synthesis and structure-property tuning of N[5]HDI oligomers is an important step toward incorporating helicenes into polymers with enhanced mechanical and electronic properties.

Continuous tracking of cortisol is central to understanding human stress physiology, yet direct electrochemical detection without biological receptors remains challenging despite its promise for stable and dynamic sensing. Electrochemical reduction of cortisol typically occurs at highly negative potentials, where parasitic interfacial currents, hydrogen evolution, and substantial capacitive background overlap with the cortisol reduction signal, preventing accurate quantification. We overcome this barrier through an integrated material–measurement strategy combining a defect-engineered graphene interface with a transition-resolved interrogation (TRI) measurement strategy. First, we engineered polybenzimidazole-derived laser-induced graphene containing nitrogen-rich defects while suppressing oxygen-derived functionalities which reduced parasitic background currents within the same cathodic potential regime. Next, we designed TRI to leverage differences in the time-dependent evolution of overlapping cathodic processes to isolate a localized cortisol-associated electrochemical reduction transition. Derivative-domain projection coupled with background estimation enables its reliable quantification. This integrated sensing architecture enables sensitive, selective, and dynamic cortisol detection in both artificial and biological interstitial fluids at low nanomolar concentrations. The response remains reproducible across physiologically relevant variations in pH, ionic strength, temperature, and repeated cycling. Together, these capabilities provide a basis for continuous electrochemical cortisol monitoring for future study of stress physiology.

The helicene N[5]HDI is selectively dibrominated, enabling a variety of cross-coupling reactions, including polycondensations. The resultant π-extended helices and polymers exhibit increased electronic delocalization upon chain lengthening.

Lithium metal batteries (LMBs) offer high energy density, but interfacial charge-transfer kinetics remain a bottleneck for high-rate operation. In this work, we construct a library of 12 ether solvents with systematic variations in symmetry and fluorination, confirming that molecular asymmetry is an important feature that results in enhanced Li+ redox kinetics and more stable solid electrolyte interphases on Li0. Furthermore, we developed a mechanistic understanding of the underlying process. We show that more tilted dipoles and lower moments of inertia strongly correlate with higher exchange current densities. Among the solvents in our study, the asymmetric F5MPE (1-methoxy-2-(2,2,3,3,3-pentafluoropropoxy) ethane) molecule, a previously unreported solvent molecule, enabled >170 stable cycles in high-rate Li||NMC full cells, outperforming the previously reported best single-salt-single-solvent ether solvent, F5DEE molecule. Overall, this work links molecular-level information with the behavior at electrochemical interfaces and provides an understanding of molecular design considerations for next-generation electrolytes tailored for high-power, fast-charging LMBs.