Search for: All records

Abstract Traumatic muscle injuries associated with volumetric muscle loss (VML) are characterized by muscle loss beyond intrinsic regeneration capacity, leading to permanent functional impairment. Experimental therapies to augment muscle regeneration, such as cell injection, are limited by low cell transplantation capacity, whereas conventional engineered muscle tissue transplants lack geometric customization to conform to the shape of the muscle defect. Here, a facile approach to engineer scaffold‐free high‐density muscle tissues in customizable geometric shapes and sizes with high cell viability and integration potential is developed. Using a facile mold‐based approach to engineer scaffold‐free modular units, transcriptional profiling is performed to uncover the role of pre‐formed cell–cell interactions within scaffold‐free muscle bioconstructs on myogenesis, an the efficacy of muscle bioconstructs in a mouse model of VML is then evaluated. RNA sequencing revealed that pre‐formed cell–cell interactions supported myogenic pathways related to muscle contraction and myofibril assembly, unlike dissociated monodisperse cells. This work further demonstrates the therapeutic efficacy of 3D rectangular solid‐shaped scaffold‐free transplants in improving muscle function and vascular regeneration. Finally, toward clinical translation, the feasibility of this technology to integrate with medical imaging and artificial intelligence‐driven customized bioconstruct design and assembly for intraoperative use is illustrated. 

In recent decades, there has been an explosion of data streams spanning the entire spectrum of biomedicine, opening novel opportunities to tackle biological and medical research questions, increasing our ability to provide effective and efficient health care. In parallel, augmented computational power has allowed the development and deployment of quantitative approaches at unprecedented scales. To effectively take advantage of this progress, it is important to invest in the training of a new generation of biomedical data scientists. Designing a graduate curriculum in the backdrop of a rapidly changing landscape of data, methods, and computing power demands flexibility and openness to adaptation. At the same time, we strive to ensure that the students acquire foundational competencies that might fuel productive and evolving careers, without being constrained to and defined by a niche trendy topic. We offer here a view of graduate training in biomedical data science from the standpoint of our experience at Stanford University. We conclude with a series of open challenges, the answers to which we believe will shape training in biomedical data science. 

Abstract Gravity waves (GWs) make crucial contributions to the middle atmospheric circulation. Yet, their climate model representation remains inaccurate, leading to key circulation biases. This study introduces a set of three neural networks (NNs) that learn to predict GW fluxes (GWFs) from multiple years of high‐resolution ERA5 reanalysis. The three NNs: a ANN, a ANN‐CNN, and an Attention UNet embed different levels of horizontal nonlocality in their architecture and are capable of representing nonlocal GW effects that are missing from current operational GW parameterizations. The NNs are evaluated offline on both time‐averaged statistics and time‐evolving flux variability. All NNs, especially the Attention UNet, accurately recreate the global GWF distribution in both the troposphere and the stratosphere. Moreover, the Attention UNet most skillfully predicts the transient evolution of GWFs over prominent orographic and nonorographic hotspots, with the model being a close second. Since even ERA5 does not resolve a substantial portion of GWFs, this deficiency is compensated by subsequently applying transfer learning on the ERA5‐trained ML models for GWFs from a 1.4 km global climate model. It is found that the re‐trained models both (a) preserve their learning from ERA5, and (b) learn to appropriately scale the predicted fluxes to account for ERA5's limited resolution. Our results highlight the importance of embedding nonlocal information for a more accurate GWF prediction and establish strategies to complement abundant reanalysis data with limited high‐resolution data to develop machine learning‐driven parameterizations for missing mesoscale processes in climate models. 

Abstract High‐power large‐aperture radar instruments observe numerous meteor head echoes per minute. Head echoes result from reflections of radio waves from plasma surrounding meteoroids as they enter Earth's atmosphere. Knowledge of the spatial distribution of electrons in this plasma is essential to determining the mass loss rate of the meteor as a function of its measured radar cross‐section. Prior work applies theoretical and computational methods to determine the electron density distribution, but assumes the meteoroid emits neutral particles uniformly across its surface. In this paper, a numerical surface ablation model demonstrates that meteoroid mass loss may occur preferentially in the direction facing the oncoming atmosphere. Specifically, meteoroid mass loss becomes proportional to the frontal surface area facing the freestream atmosphere in the limit of high Biot number, but remains isotropic in the limit of low Biot number. Meteoroid rotation has a small effect on the direction of ejected mass, but the effect is insignificant compared to variation in meteoroid properties that affect the Biot number. This result informs our computational meteor plasma model, in which we compare the effect of meteoroid vaporization on the plasma distribution in the limits of low versus high Biot number. The resulting electron density profiles demonstrate order‐of‐magnitude agreement between each other, with peak difference of 70% immediately upstream of the meteoroid. This implies that the directional distribution of vaporizing neutrals likely does not significantly influence head echo observations, lending credence to existing work that assumes isotropic ablation. 

Abstract Antiferroelectrics are a promising class of materials for applications in capacitive energy storage and multi‐state memory, but comprehensive control of their functional properties requires further research. In thin films, epitaxial strain and size effects are important tuning knobs but difficult to probe simultaneously due to low critical thicknesses of common lead‐based antiferroelectrics. Antiferroelectric NaNbO₃enables opportunities for studying size effects under strain, but electrical properties of ultra‐thin films have not been thoroughly investigated due to materials challenges. Here, high‐quality, epitaxial, coherently‐strained NaNbO₃films are synthesized from 35‐ to 250‐ nm thickness, revealing a transition from a fully ferroelectric state to coexisting ferroelectric and antiferroelectric phases with increasing thickness. The electrical performance of this phase coexistence is analyzed through positive‐up negative‐down and first‐order reversal curve measurements. Further increasing thickness leads to a fully ferroelectric state due to a strain relief mechanism that suppresses the antiferroelectricity. The potential of engineering competing ferroic orders in NaNbO₃for multiple applications is evaluated, reporting significantly enhanced recoverable energy density (20.6 J cm⁻³at 35 nm) and energy efficiency (90% at 150 nm) relative to pure bulk NaNbO₃as well as strong retention and fatigue performance with multiple accessible polarization states in the intermediate thickness films. 

Abstract Lunar paleomagnetic studies have identified multidomain metallic Fe–Ni alloys as the dominant magnetic contributors in mare basalts. Here, we explore the low‐temperature magnetic behavior of standard samples for a suite of opaque minerals that occur within mare basalts (single‐domain and multidomain Fe, wüstite, ulvöspinel, iron chromite, ilmenite, and troilite). We compare the observed low‐temperature behaviors to those of several Apollo mare basalt samples (10003, 10044, 10020, 10069, 10071, 12009, 12022, 15597). Notable magnetic transitions were detected at 30 K (ilmenite), 60–80 K (chromite, troilite), and 100–125 K (ulvöspinel, chromite). We also investigated the effects of low‐temperature cycling on mare basalt remanence and observed that only grains with coercivities 20–40 mT were cleaned. This suggests a minimal impact of diurnal temperature cycling at the lunar surface on the retrieved lunar paleointensity values. Using comprehensive electron microscopy techniques, including scanning electron microscopy (SEM), energy dispersive spectroscopy (EDS), wavelength dispersive spectroscopy (WDS), x‐ray diffraction, and transmission electron microscopy (TEM), we further examined magnetic phases within four Apollo 11 mare basalt samples. Our findings revealed the presence of Fe grains (one to 10 μm in diameter) associated with troilite contain sub‐grains ranging in size from tens to hundreds of nanometers in some samples. These grains, which fall within the single‐domain to multi‐domain range as observed in their first‐order reversal curves, might have the potential to retain high coercivity components and thereby effectively record an ancient dynamo field. 

Abstract Global climate models parameterize a range of atmospheric‐oceanic processes, including gravity waves (GWs), clouds, moist convection, and turbulence, that cannot be sufficiently resolved. These subgrid‐scale closures for unresolved processes are a substantial source of model uncertainty. Here, we present a new approach to developing machine learning (ML) parameterizations of small‐scale climate processes by fine‐tuning a pre‐trained AI foundation model (FM). FMs are largely unexplored in climate research. A pre‐trained encoder‐decoder from a 2.3 billion parameter FM (NASA and IBM Research's Prithvi WxC)—which contains a latent probabilistic representation of atmospheric evolution—is fine‐tuned (or reused) to create a deep learning parameterization for atmospheric gravity waves (GWs); a process unseen during pre‐training. The parameterization captures GW effects for a coarse‐resolution climate model by learning the fluxes from an atmospheric reanalysis with 10 times finer resolution. A comparison of monthly averages and instantaneous evolution with a machine learning model baseline (an Attention U‐Net) reveals superior predictive performance of the FM parameterization throughout the atmosphere, even in regions excluded during pre‐training. This performance boost is quantified using the Hellinger distance, which is 0.11 for the baseline and 0.06 for the fine‐tuned model. Our findings emphasize the versatility and reusability of FMs, which could be used to accomplish a range of atmosphere‐ and climate‐related applications, leading the way for the creation of observations‐driven and physically accurate parameterizations for more earth system processes.