Materials design for hypersonics.

Hypersonic vehicles must withstand extreme conditions during flights that exceed five times the speed of sound. These systems have the potential to facilitate rapid access to space, bolster defense capabilities, and create a new paradigm for transcontinental earth-to-earth travel. However, extreme aerothermal environments create significant challenges for vehicle materials and structures. This work addresses the critical need to develop resilient refractory alloys, composites, and ceramics. We will highlight key design principles for critical vehicle areas such as primary structures, thermal protection, and propulsion systems; the role of theory and computation; and strategies for advancing laboratory-scale materials to manufacturable flight-ready components.

Matrix confinement modulates 3D spheroid sorting and burst-like collective migration.

While it is known that cells with differential adhesion tend to segregate and preferentially sort, the physical forces governing sorting and invasion in heterogeneous tumors remain poorly understood. To investigate this, we tune matrix confinement, mimicking changes in the stiffness and confinement of the tumor microenvironment, to explore how physical confinement influences individual and collective cell migration in 3D spheroids. High levels of confinement lead to cell sorting while reducing matrix confinement triggers the collective fluidization of cell motion. Cell sorting, which depends on cell-cell adhesion, is crucial to this phenomenon. Burst-like migration does not occur for spheroids that have not undergone sorting, regardless of the degree of matrix confinement. Using computational Self-Propelled Voronoi modeling, we show that spheroid sorting and invasion into the matrix depend on the balance between cell-generated forces and matrix resistance. The findings support a model where matrix confinement modulates 3D spheroid sorting and unjamming in an adhesion-dependent manner, providing insights into the mechanisms of cell sorting and migration in the primary tumor and toward distant metastatic sites. STATEMENT OF SIGNIFICANCE: The mechanical properties of the tumor microenvironment significantly influence cancer cell migration within the primary tumor, yet how these properties affect intercellular interactions in heterogeneous tumors is not well understood. By utilizing calcium and calcium chelators, we dynamically alter collagen-alginate hydrogel stiffness and investigate tumor cell behavior within co-culture spheroids in response to varying degrees of matrix confinement. High confinement is found to trigger cell sorting while reducing confinement for sorted spheroids facilitates collective cell invasion. Notably, without prior sorting, spheroids do not exhibit burst-like migration, regardless of confinement levels. This work establishes that matrix confinement and intercellular adhesion regulate 3D spheroid dynamics, offering insights into cellular organization and migration within the primary tumor.

Cellular mechanotransduction of human osteoblasts in microgravity.

Astronauts experience significant and rapid bone loss as a result of an extended stay in space, making the International Space Station (ISS) the perfect laboratory for studying osteoporosis due to the accelerated nature of bone loss on the ISS. This prompts the question, how does the lack of load due to zero-gravity propagate to bone-forming cells, human fetal osteoblasts (hFOBs), altering their maturation to mineralization? Here, we aim to study the mechanotransduction mechanisms by which bone loss occurs in microgravity. Two automated experiments, 4 microfluidic chips capable of measuring single-cell mechanics of hFOBs via aspiration and cell spheroids incubated in pressure-controlled chambers, were each integrated into a CubeLab deployed to the ISS National Laboratory. For the first experiment, we report protrusion measurements of aspirated cells after exposure to microgravity at the ISS and compare these results to ground control conducted inside the CubeLab. Our analysis revealed slightly elongated protrusions for space samples compared to ground samples indicating softening of hFOB cells in microgravity. In the second experiment, we encapsulated osteoblast spheroids in collagen gel and incubated the samples in pressure-controlled chambers. We found that microgravity significantly reduced filamentous actin levels in the hFOB spheroids. When subjected to pressure, the spheroids exhibited increased pSMAD1/5/9 expression, regardless of the microgravity condition. Moreover, microgravity reduced YAP expression, while pressure increased YAP levels, thus restoring YAP expression for spheroids in microgravity. Our study provides insights into the influence of microgravity on the mechanical properties of bone cells and the impact of compressive pressure on cell behavior and signaling in space.

Cellular mechanotransduction of human osteoblasts in microgravity.

Astronauts experience significant and rapid bone loss as a result of an extended stay in space, making the International Space Station (ISS) the perfect laboratory for studying osteoporosis due to the accelerated nature of bone loss on the ISS. This prompts the question, how does the lack of load due to zero-gravity propagate to bone-forming cells, human fetal osteoblasts (hFOBs), altering their maturation to mineralization? Here, we aim to study the mechanotransduction mechanisms by which bone loss occurs in microgravity. Two automated experiments, microfluidic chips capable of measuring single-cell mechanics via aspiration and cell spheroids incubated in pressure-controlled chambers, were each integrated into a CubeLab deployed to the ISS National Laboratory. For the first experiment, we report protrusion measurements of aspirated cells after exposure to microgravity at the ISS and compare these results to ground control conducted inside the CubeLab. We found slightly elongated protrusions for space samples compared to ground samples indicating softening of hFOB cells in microgravity. In the second experiment, we encapsulated osteoblast spheroids in collagen gel and incubated the samples in pressure-controlled chambers. We found that microgravity significantly reduced filamentous actin levels in the hFOB spheroids. When subjected to pressure, the spheroids exhibited increased pSMAD1/5/9 expression, regardless of the microgravity condition. Moreover, microgravity reduced YAP expression, while pressure increased YAP levels, thus restoring YAP expression for spheroids in microgravity. Our study provides insights into the influence of microgravity on the mechanical properties of bone cells and the impact of compressive pressure on cell signaling in space.

An Op-Ed Writing Curriculum for Medical Students to Engage in Advocacy Through Public Writing.

BACKGROUND: Op-ed writing can be a powerful and accessible advocacy tool for physicians, but training is lacking in undergraduate medical education. AIM: To train and engage first-year medical students in op-ed writing. SETTING: Midwestern research-intensive medical school. PARTICIPANTS: All students in a required first-year health policy course in 2021 and 2022. PROGRAM DESCRIPTION: For their health policy course's final assignment, students could opt to write an op-ed on a healthcare issue of their choice. All students received written instruction on op-ed writing. Additionally, they could access a seminar, coaching and editing by peers and faculty, and publication guidance. PROGRAM EVALUATION: Of 179 students over 2 years, 105 chose to write op-eds. Fifty-one attended the seminar, 35 attended peer coaching sessions, 33 accessed structured peer editing, and 23 received faculty assistance. Thirty-eight students submitted a total of 42 op-eds for publication. Twenty-two pieces were published in major outlets and 17 in the university's health policy review. Of the 22 in major outlets, 21 received editing from either peers or faculty. DISCUSSION: An op-ed writing curriculum can be integrated into an existing medical school health policy course, resulting in a high level of engagement and in published op-eds by medical students.

Development of mechanosensitive synthetic cells for biomedical applications.

The ability of cells to sense and respond to their physical environment plays a fundamental role in a broad spectrum of biological processes. As one of the most essential molecular force sensors and transducers found in cell membranes, mechanosensitive (MS) ion channels can convert mechanical inputs into biochemical or electrical signals to mediate a variety of sensations. The bottom-up construction of cell-sized compartments displaying cell-like organization, behaviors, and complexity, also known as synthetic cells, has gained popularity as an experimental platform to characterize biological functions in isolation. By reconstituting MS channels in the synthetic lipid bilayers, we envision using mechanosensitive synthetic cells for several medical applications. Here, we describe three different concepts for using ultrasound, shear stress, and compressive stress as mechanical stimuli to activate drug release from mechanosensitive synthetic cells for disease treatments.

Matrix confinement modulates 3D spheroid sorting and burst-like collective migration.

While it is known that cells with differential adhesion tend to segregate and preferentially sort, the physical forces governing sorting and invasion in heterogeneous tumors remain poorly understood. To investigate this, we tune matrix confinement, mimicking changes in the stiffness and confinement of the tumor microenvironment, to explore how physical confinement influences individual and collective cell migration in 3D spheroids. High levels of confinement lead to cell sorting while reducing matrix confinement triggers the collective fluidization of cell motion. Cell sorting, which depends on cell-cell adhesion, is crucial to this phenomenon. Burst-like migration does not occur for spheroids that have not undergone sorting, regardless of the degree of matrix confinement. Using computational Self-Propelled Voronoi modeling, we show that spheroid sorting and invasion into the matrix depend on the balance between cell-generated forces and matrix resistance. The findings support a model where matrix confinement modulates 3D spheroid sorting and unjamming in an adhesion-dependent manner, providing insights into the mechanisms of cell sorting and migration in the primary tumor and toward distant metastatic sites.

Technique and Outcomes of Concomitant Aortic and Caval Resection and Reconstruction for Retroperitoneal Tumors.

BACKGROUND: Major vascular involvement is often considered a contraindication to resection of malignant tumors, but in highly selected patients, it can be performed safely, with results that are highly dependent upon the tumor biology. Resection of both the aorta and inferior vena cava (IVC) is a rare undertaking, requiring both favorable tumor biology and a patient fit for a substantial surgical insult; nevertheless, it provides the possibility of a cure. METHODS: Patients requiring resection and reconstruction of both the aorta and IVC from 2009 through 2019 at 2 university medical centers were included. Patient characteristics, operative technique, and outcomes were retrospectively collected. RESULTS: We identified 9 patients, all with infrarenal reconstruction or repair of the aorta and IVC. All cases were performed with systemic heparinization and required simultaneous aortic and caval cross-clamping for tumor resection. No temporary venous or arterial bypass was used. Since arterial reperfusion with the IVC clamped was poorly tolerated in 1 patient, venous reconstruction was typically completed first. Primary repair was performed in 1 patient, while 8 required replacements. In 2 patients, aortic homograft was used for replacement of both the aortoiliac and iliocaval segments in contaminated surgical fields. In the remaining 6, Dacron was used for arterial replacement; either Dacron (n = 2) or polytetrafluoroethylene (n = 4) were used for venous replacement. Patients were discharged after a median stay of 8 days (range: 5-16). At median follow-up of 17 months (range 3-79 months), 2 patients with paraganglioma and 1 patient with Leydig cell carcinoma had cancer recurrences. Venous reconstructions occluded in 3 patients (38%), although symptoms were minimal. One patient presented acutely with a thrombosed iliac artery limb and bilateral common iliac artery anastomotic stenoses, treated successfully with thrombolysis and stenting. CONCLUSIONS: Patients with tumor involving both the aorta and IVC can be successfully treated with resection and reconstruction. En bloc tumor resection, restoration of venous return before arterial reconstruction, and most importantly, careful patient selection, all contribute to positive outcomes in this otherwise incurable population.

Optimizing quantum noise-induced reservoir computing for nonlinear and chaotic time series prediction.

Quantum reservoir computing is strongly emerging for sequential and time series data prediction in quantum machine learning. We make advancements to the quantum noise-induced reservoir, in which reservoir noise is used as a resource to generate expressive, nonlinear signals that are efficiently learned with a single linear output layer. We address the need for quantum reservoir tuning with a novel and generally applicable approach to quantum circuit parameterization, in which tunable noise models are programmed to the quantum reservoir circuit to be fully controlled for effective optimization. Our systematic approach also involves reductions in quantum reservoir circuits in the number of qubits and entanglement scheme complexity. We show that with only a single noise model and small memory capacities, excellent simulation results were obtained on nonlinear benchmarks that include the Mackey-Glass system for 100 steps ahead in the challenging chaotic regime.

Calcium-triggered DNA-mediated membrane fusion in synthetic cells.

In cells, membrane fusion is mediated by SNARE proteins, whose activities are calcium-dependent. While several non-native membrane fusion mechanisms have been demonstrated, few can respond to external stimuli. Here, we develop a calcium-triggered DNA-mediated membrane fusion strategy where fusion is regulated using surface-bound PEG chains that are cleavable by the calcium-activated protease calpain-1.

Comparing genomic and epigenomic features across species using the WashU Comparative Epigenome Browser.

Genome browsers have become an intuitive and critical tool to visualize and analyze genomic features and data. Conventional genome browsers display data/annotations on a single reference genome/assembly; there are also genomic alignment viewer/browsers that help users visualize alignment, mismatch, and rearrangement between syntenic regions. However, there is a growing need for a comparative epigenome browser that can display genomic and epigenomic data sets across different species and enable users to compare them between syntenic regions. Here, we present the WashU Comparative Epigenome Browser. It allows users to load functional genomic data sets/annotations mapped to different genomes and display them over syntenic regions simultaneously. The browser also displays genetic differences between the genomes from single-nucleotide variants (SNVs) to structural variants (SVs) to visualize the association between epigenomic differences and genetic differences. Instead of anchoring all data sets to the reference genome coordinates, it creates independent coordinates of different genome assemblies to faithfully present features and data mapped to different genomes. It uses a simple, intuitive genome-align track to illustrate the syntenic relationship between different species. It extends the widely used WashU Epigenome Browser infrastructure and can be expanded to support multiple species. This new browser function will greatly facilitate comparative genomic/epigenomic research, as well as support the recent growing needs to directly compare and benchmark the T2T CHM13 assembly and other human genome assemblies.

Calcium-triggered DNA-mediated membrane fusion in synthetic cells.

In cells, membrane fusion is mediated by SNARE proteins, whose activities are calcium-dependent. While several non-native membrane fusion mechanisms have been demonstrated, few can respond to external stimuli. Here, we develop a calcium-triggered DNA-mediated membrane fusion strategy where fusion is regulated using surface-bound PEG chains that are cleavable by the calcium-activated protease calpain-1.

"I am not a junkie": Social categorization and differentiation among people who use drugs.

BACKGROUND: Substance use stigma is a form of group-based exclusion, and delineating pathways from stigma to poor health requires a deeper understanding of the social dynamics of people who use drugs (PWUD). Outside of recovery, scant research has examined the role of social identity in addiction. Framed by Social Identity Theory/Self-Categorization Theory, this qualitative study investigated strategies of within-group categorization and differentiation among PWUD and the roles these social categories may play in shaping intragroup attitudes, perceptions, and behaviors. METHODS: Data come from the Rural Opioid Initiative, a multi-site study of the overdose epidemic in rural United States. We conducted in-depth interviews with people who reported using opioids or injecting any drug (n=355) living in 65 counties across 10 states. Interviews focused on participants' biographical histories, past and current drug use, risk behaviors, and experiences with healthcare providers and law enforcement. Social categories and dimensions along which categories were evaluated were inductively identified using reflexive thematic analysis. RESULTS: We identified seven social categories that were commonly appraised by participants along eight evaluative dimensions. Categories included drug of choice, route of administration, method of attainment, gender, age, genesis of use, and recovery approach. Categories were evaluated by participants based on ascribed characteristics of morality, destructiveness, aversiveness, control, functionality, victimhood, recklessness, and determination. Participants performed nuanced identity work during interviews, including reifying social categories, defining 'addict' prototypicality, reflexively comparing self to other, and disidentifying from the PWUD supra-category. CONCLUSION: We identify several facets of identity, both behavioral and demographic, along which people who use drugs perceive salient social boundaries. Beyond an addiction-recovery binary, identity is shaped by multiple aspects of the social self in substance use. Patterns of categorization and differentiation revealed negative intragroup attitudes, including stigma, that may hinder solidary-building and collective action in this marginalized group.

BrainGENIE: The Brain Gene Expression and Network Imputation Engine.

In vivo experimental analysis of human brain tissue poses substantial challenges and ethical concerns. To address this problem, we developed a computational method called the Brain Gene Expression and Network-Imputation Engine (BrainGENIE) that leverages peripheral-blood transcriptomes to predict brain tissue-specific gene-expression levels. Paired blood-brain transcriptomic data collected by the Genotype-Tissue Expression (GTEx) Project was used to train BrainGENIE models to predict gene-expression levels in ten distinct brain regions using whole-blood gene-expression profiles. The performance of BrainGENIE was compared to PrediXcan, a popular method for imputing gene expression levels from genotypes. BrainGENIE significantly predicted brain tissue-specific expression levels for 2947-11,816 genes (false-discovery rate-adjusted p < 0.05), including many transcripts that cannot be predicted significantly by a transcriptome-imputation method such as PrediXcan. BrainGENIE recapitulated measured diagnosis-related gene-expression changes in the brain for autism, bipolar disorder, and schizophrenia better than direct correlations from blood and predictions from PrediXcan. We developed a convenient software toolset for deploying BrainGENIE, and provide recommendations for how best to implement models. BrainGENIE complements and, in some ways, outperforms existing transcriptome-imputation tools, providing biologically meaningful predictions and opening new research avenues.

Quantum machine learning with differential privacy.

Quantum machine learning (QML) can complement the growing trend of using learned models for a myriad of classification tasks, from image recognition to natural speech processing. There exists the potential for a quantum advantage due to the intractability of quantum operations on a classical computer. Many datasets used in machine learning are crowd sourced or contain some private information, but to the best of our knowledge, no current QML models are equipped with privacy-preserving features. This raises concerns as it is paramount that models do not expose sensitive information. Thus, privacy-preserving algorithms need to be implemented with QML. One solution is to make the machine learning algorithm differentially private, meaning the effect of a single data point on the training dataset is minimized. Differentially private machine learning models have been investigated, but differential privacy has not been thoroughly studied in the context of QML. In this study, we develop a hybrid quantum-classical model that is trained to preserve privacy using differentially private optimization algorithm. This marks the first proof-of-principle demonstration of privacy-preserving QML. The experiments demonstrate that differentially private QML can protect user-sensitive information without signficiantly diminishing model accuracy. Although the quantum model is simulated and tested on a classical computer, it demonstrates potential to be efficiently implemented on near-term quantum devices [noisy intermediate-scale quantum (NISQ)]. The approach's success is illustrated via the classification of spatially classed two-dimensional datasets and a binary MNIST classification. This implementation of privacy-preserving QML will ensure confidentiality and accurate learning on NISQ technology.

Quality of care and prescription patterns among patients with diabetic kidney disease-a large-scale cohort study from Taiwanese clinics.

Aims: To investigate the quality of care and prescription patterns of patients with diabetic kidney disease (DKD) receiving primary care at local clinics in Taiwan. Methods: A retrospective chart review was conducted in 43 primary care clinics in Taiwan. The patients' baseline characteristics, laboratory tests, presence of complications and antidiabetic agents prescribed were analyzed. Results: 7,200 patients with type 2 diabetes mellitus were enrolled. Percentage of HbA1c, blood pressure (BP), and low density lipoprotein cholesterol (LDL-C) goals reached were 52.5% in HbA1c < 7%, 40.9% in BP < 130/80 mmHg and 79.7% in LDL-C < 2.59 mmol/L. 18.3% achieved all three ABC goals. However, patients with DKD had a lower rate of ABC goal attainment and higher rate of complications. Among DKD patients with eGFR ≥ 30 ml/min/1.73 m2 and on monotherapy, metformin was most frequently prescribed. As for dual therapy, the most common combinations were metformin with sulfonylurea and metformin with DPP-4 inhibitors. Conclusions: Diabetes patients in Taiwan receiving primary diabetes care at local clinics had generally satisfactory management performance. However, more aggressive HbA1c, BP, and LDL-C management among DKD patients should be emphasized. Contrary to current recommendations, SGLT-2 inhibitors and GLP-1 receptor agonists as frontline therapy were under-prescribed.

Management of acute aortoiliac arterial thrombosis in patients with the novel coronavirus disease 2019: A case series and systematic review of the literature.

Objectives: Venous thrombosis has been widely described in the setting of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection; however, arterial thrombosis has rarely been reported. This study aims to assess the incidence, risk factors, interventions, and outcomes of acute aortoiliac arterial thrombosis in patients with active SARS-CoV-2 infections. Methods: We present seven SARS-CoV-2-positive patients from our institution who acutely developed thrombi in the aortoiliac arterial system (7/2020-1/2021). A systematic review of the literature on aortoiliac arterial thrombosis in patients with SARS-CoV-2 infections in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines was also performed. The available data from all reported cases in the literature and at our institution were analyzed. Results: Thirty published articles and journal correspondences, including 52 patients, were reviewed and analyzed in addition to our institution's 7 cases. In total, 59 SARS-CoV-2-positive patients were found to have acute aortoiliac thrombosis. The abdominal aorta was the most frequent location for the development of a thrombus. Baseline demographics and medical comorbidities were not significantly different between the symptomatic and asymptomatic cohorts. Seventy-one percent of patients were symptomatic (lower limb ischemia: 75.0%, renal infarction: 20.0%, stroke: 12.5%, mesenteric ischemia: 10.0%). All patients with thrombus involving the ascending aorta, aortic bifurcation, or iliac artery developed thromboembolic or ischemic complications. All patients received systemic anticoagulation. Fifty-three percent of all patients were managed medically. Ninety-four percent of the asymptomatic patients were managed medically. One asymptomatic patient underwent endovascular aspiration of a mobile thrombus. Three (23.1%) deaths occurred in the asymptomatic cohort from hypoxic respiratory failure. Fourteen (36.8%) deaths occurred in the symptomatic cohort. The in-hospital mortality rate was 33.3% overall and 43.8% for patients with thrombi involving more than one aortoiliac segment. Conclusions: The presence of thrombi in the aortoiliac arterial system appears to be a poor prognostic indicator for patients with active SARS-CoV-2 infections. Medical management of patients with asymptomatic aortoiliac thrombi may be considered. The presence of thrombi involving the ascending aorta, aortic bifurcation, or iliac artery may warrant consideration for operative intervention due to the risk for thromboembolic or ischemic complications. Further study is needed to fully delineate the risk factors, optimal treatment, and outcomes of arterial thrombosis in the setting of SARS-CoV-2 infection.

WashU Epigenome Browser update 2022.

WashU Epigenome Browser (https://epigenomegateway.wustl.edu/browser/) is a web-based genomic data exploration tool that provides visualization, integration, and analysis of epigenomic datasets. The newly renovated user interface and functions have enabled researchers to engage with the browser and genomic data more efficiently and effectively since 2018. Here, we introduce a new integrated panel design in the browser that allows users to interact with 1D (genomic features), 2D (such as Hi-C), 3D (genome structure), and 4D (time series) data in a single web page. The browser can display three-dimensional chromatin structures with the 3D viewer module. The 4D tracks, called 'Dynamic' tracks, animatedly display time-series data, allowing for a more striking visual impact to identify the gene or genomic region candidates as a function of time. Genomic data, such as annotation features, numerical values, and chromatin interaction data can all be viewed in the dynamic track mode. Imaging data from microscopy experiments can also be displayed in the browser. In addition to software development, we continue to service and expand the data hubs we host for large consortia including 4DN, Roadmap Epigenomics, TaRGET and ENCODE, among others. Our growing user/developer community developed additional track types as plugins, such as qBed and dynseq tracks, which extend the utility of the browser. The browser serves as a foundation for additional genomics platforms including the WashU Virus Genome Browser (for COVID-19 research) and the Comparative Genome Browser. The WashU Epigenome Browser can also be accessed freely through Amazon Web Services at https://epigenomegateway.org/.

Summix: A method for detecting and adjusting for population structure in genetic summary data.

Publicly available genetic summary data have high utility in research and the clinic, including prioritizing putative causal variants, polygenic scoring, and leveraging common controls. However, summarizing individual-level data can mask population structure, resulting in confounding, reduced power, and incorrect prioritization of putative causal variants. This limits the utility of publicly available data, especially for understudied or admixed populations where additional research and resources are most needed. Although several methods exist to estimate ancestry in individual-level data, methods to estimate ancestry proportions in summary data are lacking. Here, we present Summix, a method to efficiently deconvolute ancestry and provide ancestry-adjusted allele frequencies (AFs) from summary data. Using continental reference ancestry, African (AFR), non-Finnish European (EUR), East Asian (EAS), Indigenous American (IAM), South Asian (SAS), we obtain accurate and precise estimates (within 0.1%) for all simulation scenarios. We apply Summix to gnomAD v.2.1 exome and genome groups and subgroups, finding heterogeneous continental ancestry for several groups, including African/African American (∼84% AFR, ∼14% EUR) and American/Latinx (∼4% AFR, ∼5% EAS, ∼43% EUR, ∼46% IAM). Compared to the unadjusted gnomAD AFs, Summix's ancestry-adjusted AFs more closely match respective African and Latinx reference samples. Even on modern, dense panels of summary statistics, Summix yields results in seconds, allowing for estimation of confidence intervals via block bootstrap. Given an accompanying R package, Summix increases the utility and equity of public genetic resources, empowering novel research opportunities.

A seq2seq model to forecast the COVID-19 cases, deaths and reproductive R numbers in US counties.

The global pandemic of coronavirus disease 2019 (COVID-19) has killed almost two million people worldwide and over 400 thousand in the United States (US). As the pandemic evolves, informed policy-making and strategic resource allocation relies on accurate forecasts. To predict the spread of the virus within US counties, we curated an array of county-level demographic and COVID-19-relevant health risk factors. In combination with the county-level case and death numbers curated by John Hopkins university, we developed a forecasting model using deep learning (DL). We implemented an autoencoder-based Seq2Seq model with gated recurrent units (GRUs) in the deep recurrent layers. We trained the model to predict future incident cases, deaths and the reproductive number, R For most counties, it makes accurate predictions of new incident cases, deaths and R values, up to 30 days in the future. Our framework can also be used to predict other targets that are useful indices for policymaking, for example hospitalization or the occupancy of intensive care units. Our DL framework is publicly available on GitHub and can be adapted for other indices of the COVID-19 spread. We hope that our forecasts and model can help local governments in the continued fight against COVID-19.