Search

Background Cystic fibrosis (CF) is a genetic disease caused by mutations in the gene encoding the cystic fibrosis transmembrane conductance regulator (CFTR), which results in impaired airway mucociliary clearance, inflammation, infection, and respiratory insufficiency. The development of new therapeutics for CF are limited by the lack of reliable preclinical models that recapitulate the structural, immunological, and bioelectrical features of human CF lungs. Methods We leveraged organ-on-a-chip technology to develop a microfluidic device lined by primary human CF bronchial epithelial cells grown under an air-liquid interface and interfaced with pulmonary microvascular endothelial cells (CF Airway Chip) exposed to fluid flow. The responses of CF and healthy Airway Chips were analyzed in the presence or absence of polymorphonuclear leukocytes (PMNs) and the bacterial pathogen, Pseudomonas aeruginosa. Results The CF Airway Chip faithfully recapitulated many features of the human CF airways, including enhanced mucus accumulation, increased cilia density, and a higher ciliary beating frequency compared to chips lined by healthy bronchial epithelial cells. The CF chips also secreted higher levels of IL-8, which was accompanied by enhanced PMN adhesion to the endothelium and transmigration into the airway compartment. In addition, CF Airway Chips provided a more favorable environment for Pseudomonas aeruginosa growth, which resulted in enhanced secretion of inflammatory cytokines and recruitment of PMNs to the airway. Conclusions The human CF Airway Chip may provide a valuable preclinical tool for pathophysiology studies as well as for drug testing and personalized medicine.

In this work we explore confidence elicitation methods for crowdsourcing "soft" labels, e.g., probability estimates, to reduce the annotation costs for domains with ambiguous data. Machine learning research has shown that such "soft" labels are more informative and can reduce the data requirements when training supervised machine learning models. By reducing the number of required labels, we can reduce the costs of slow annotation processes such as audio annotation. In our experiments we evaluated three confidence elicitation methods: 1) "No Confidence" elicitation, 2) "Simple Confidence" elicitation, and 3) "Betting" mechanism for confidence elicitation, at both individual (i.e., per participant) and aggregate (i.e., crowd) levels. In addition, we evaluated the interaction between confidence elicitation methods, annotation types (binary, probability, and z-score derived probability), and "soft" versus "hard" (i.e., binarized) aggregate labels. Our results show that both confidence elicitation mechanisms result in higher annotation quality than the "No Confidence" mechanism for binary annotations at both participant and recording levels. In addition, when aggregating labels at the recording level, results indicate that we can achieve comparable results to those with 10-participant aggregate annotations using fewer annotators if we aggregate "soft" labels instead of "hard" labels. These results suggest that for binary audio annotation using a confidence elicitation mechanism and aggregating continuous labels we can obtain higher annotation quality, more informative labels, with quality differences more pronounced with fewer participants. Finally, we propose a way of integrating these confidence elicitation methods into a two-stage, multi-label annotation pipeline.

Most bacterial vaccines work for a subset of bacterial strains or require the modification of the antigen or isolation of the pathogen before vaccine development. Here we report injectable biomaterial vaccines that trigger potent humoral and T-cell responses to bacterial antigens by recruiting, reprogramming and releasing dendritic cells. The vaccines are assembled from regulatorily approved products and consist of a scaffold with absorbed granulocyte-macrophage colony-stimulating factor and CpG-rich oligonucleotides incorporating superparamagnetic microbeads coated with the broad-spectrum opsonin Fc-mannose-binding lectin for the magnetic capture of pathogen-associated molecular patterns from inactivated bacterial-cell-wall lysates. The vaccines protect mice against skin infection with methicillin-resistant Staphylococcus aureus, mice and pigs against septic shock from a lethal Escherichia coli challenge and, when loaded with pathogen-associated molecular patterns isolated from infected animals, uninfected animals against a challenge with different E. coli serotypes. The strong immunogenicity and low incidence of adverse events, a modular manufacturing process, and the use of components compatible with current good manufacturing practice could make this vaccine technology suitable for responding to bacterial pandemics and biothreats.

The Sounds of New York City (SONYC) project (2016–2022) was a project to monitor and mitigate urban noise pollution using a smart acoustic sensor network, citizen scientists, and collaboration with city agencies. During its lifetime, the project deployed 75 fixed-location sensors and collected over 150 × 106 10-s audio recordings. A key element of the project was the development of machine listening models to detect the sources of noise pollution rather than just the overall noise level. In this talk, we first discuss our initial approach to data collection and machine listening including sensor development and deployment, citizen-science data annotation, self-supervised audio representation learning, and downstream sound-event detection model training. We then discuss analysis results using the outputs of these models, followed by the challenges and limitations of our initial approach. Finally, we discuss our solutions to overcome those challenges, such as citizen-deployed sensors, source-specific loudness estimation, and few-shot sound-event detection.

Noise is one of the primary quality-of-life issues in urban environments. In addition to annoyance, noise negatively impacts public health and educational performance. While low-cost sensors can be deployed to monitor ambient noise levels at high temporal resolutions, the amount of data they produce and the complexity of these data pose significant analytical challenges. One way to address these challenges is through machine listening techniques, which are used to extract features in attempts to classify the source of noise and understand temporal patterns of a city's noise situation. However, the overwhelming number of noise sources in the urban environment and the scarcity of labeled data makes it nearly impossible to create classification models with large enough vocabularies that capture the true dynamism of urban soundscapes In this paper, we first identify a set of requirements in the yet unexplored domain of urban soundscape exploration. To satisfy the requirements and tackle the identified challenges, we propose Urban Rhapsody, a framework that combines state-of-the-art audio representation, machine learning, and visual analytics to allow users to interactively create classification models, understand noise patterns of a city, and quickly retrieve and label audio excerpts in order to create a large high-precision annotated database of urban sound recordings. We demonstrate the tool's utility through case studies performed by domain experts using data generated over the five-year deployment of a one-of-a-kind sensor network in New York City., Comment: Accepted at EuroVis 2022. Source code available at: https://github.com/VIDA-NYU/Urban-Rhapsody

Few-shot learning aims to train models that can recognize novel classes given just a handful of labeled examples, known as the support set. While the field has seen notable advances in recent years, they have often focused on multi-class image classification. Audio, in contrast, is often multi-label due to overlapping sounds, resulting in unique properties such as polyphony and signal-to-noise ratios (SNR). This leads to unanswered questions concerning the impact such audio properties may have on few-shot learning system design, performance, and human-computer interaction, as it is typically up to the user to collect and provide inference-time support set examples. We address these questions through a series of experiments designed to elucidate the answers to these questions. We introduce two novel datasets, FSD-MIX-CLIPS and FSD-MIX-SED, whose programmatic generation allows us to explore these questions systematically. Our experiments lead to audio-specific insights on few-shot learning, some of which are at odds with recent findings in the image domain: there is no best one-size-fits-all model, method, and support set selection criterion. Rather, it depends on the expected application scenario. Our code and data are available at https://github.com/wangyu/rethink-audio-fsl., WASPAA 2021

Current therapeutic strategies against bacterial infections focus on reduction of pathogen load through antibiotics; however, stimulation of host tolerance to infection might offer an alternative approach. Here we used computational transcriptomics and a Xenopus embryo infection model to rapidly discover infection response pathways, identify potential tolerance inducer drugs, and validate their ability to induce broad tolerance. Xenopus embryos exhibit natural tolerance to A. baumanii, K. pneumoniae, S. aureus, and S. pneumoniae bacteria, whereas A. hydrophila and P. aeruginosa produce infection that leads to death. Transcriptional profiling led to definition of a 20-gene signature that allows for discrimination between tolerant and susceptible states, as well as identification of active and passive tolerance responses based on the degree of engagement of gene transcription modulation. Upregulation of metal ion transport and hypoxia pathways reminiscent of responses observed in primate and mouse infection models were identified as tolerance mediators, and drug screening in the susceptible A. hydrophila infection model confirmed that a metal chelator (deferoxamine) and HIF-1α agonist (1,4-DPCA) increase embryo survival despite high pathogen load. These data demonstrate the value of combining the Xenopus embryo infection model with multi-omics analyses for mechanistic discovery and drug repurposing to induce host tolerance to bacterial infection.

The coronavirus disease 2019 (COVID‐19) pandemic demonstrates the importance of generating safe and efficacious vaccines that can be rapidly deployed against emerging pathogens. Subunit vaccines are considered among the safest, but proteins used in these typically lack strong immunogenicity, leading to poor immune responses. Here, a biomaterial COVID‐19 vaccine based on a mesoporous silica rods (MSRs) platform is described. MSRs loaded with granulocyte‐macrophage colony‐stimulating factor (GM‐CSF), the toll‐like receptor 4 (TLR‐4) agonist monophosphoryl lipid A (MPLA), and SARS‐CoV‐2 viral protein antigens slowly release their cargo and form subcutaneous scaffolds that locally recruit and activate antigen‐presenting cells (APCs) for the generation of adaptive immunity. MSR‐based vaccines generate robust and durable cellular and humoral responses against SARS‐CoV‐2 antigens, including the poorly immunogenic receptor binding domain (RBD) of the spike (S) protein. Persistent antibodies over the course of 8 months are found in all vaccine configurations tested and robust in vitro viral neutralization is observed both in a prime‐boost and a single‐dose regimen. These vaccines can be fully formulated ahead of time or stored lyophilized and reconstituted with an antigen mixture moments before injection, which can facilitate its rapid deployment against emerging SARS‐CoV‐2 variants or new pathogens. Together, the data show a promising COVID‐19 vaccine candidate and a generally adaptable vaccine platform against infectious pathogens., Mesoporous silica rods (MSRs) scaffolds are used as a platform to induce adaptive immunity against SARS‐CoV‐2 antigens. Sustained delivery of GM‐CSF, MPLA, and a variety of SARS‐CoV‐2 subunit immunogens from the MSR vaccine recruite immune cells into the scaffolds and consistently induce antigen‐specific long‐lasting antibodies, cytotoxic T lymphocyte responses (CTL), and viral neutralization in vitro.

BackgroundCystic fibrosis (CF) is a genetic disease caused by mutations in the gene encoding the cystic fibrosis transmembrane conductance regulator (CFTR), which results in impaired airway mucociliary clearance, inflammation, infection, and respiratory insufficiency. The development of new therapeutics for CF are limited by the lack of reliable preclinical models that recapitulate the structural, immunological, and bioelectrical features of human CF lungs.MethodsWe leveraged organ-on-a-chip technology to develop a microfluidic device lined by primary human CF bronchial epithelial cells grown under an air-liquid interface and interfaced with pulmonary microvascular endothelial cells (CF Airway Chip) exposed to fluid flow. The responses of CF and healthy Airway Chips were analyzed in the presence or absence of polymorphonuclear leukocytes (PMNs) and the bacterial pathogen, Pseudomonas aeruginosa.ResultsThe CF Airway Chip faithfully recapitulated many features of the human CF airways, including enhanced mucus production, increased cilia density and a higher ciliary beating frequency compared to chips lined by healthy bronchial epithelial cells. The CF chips also secreted higher levels of IL-8, which was accompanied by enhanced PMN adhesion to the endothelium and transmigration into the airway compartment. In addition, CF Airway Chips provided a more favorable environment for Pseudomonas aeruginosa growth, which resulted in enhanced secretion of inflammatory cytokines and recruitment of PMNs to the airway.ConclusionsThe human CF Airway Chip may provide a valuable preclinical tool for pathophysiology studies as well as for drug testing and personalized medicine.

Embedding models that encode semantic information into low-dimensional vector representations are useful in various machine learning tasks with limited training data. However, these models are typically too large to support inference in small edge devices, which motivates training of smaller yet comparably predictive student embedding models through knowledge distillation (KD). While knowledge distillation traditionally uses the teacher’s original training dataset to train the student, we hypothesize that using a dataset similar to the student’s target domain allows for better compression and training efficiency for the said domain, at the cost of reduced generality across other (non-pertinent) domains. Hence, we introduce Specialized Embedding Approximation (SEA) to train a student featurizer to approximate the teacher’s embedding manifold for a given target domain. We demonstrate the feasibility of SEA in the context of acoustic event classification for urban noise monitoring and show that leveraging a dataset related to this target domain not only improves the baseline performance of the original embedding model but also yields competitive students with >1 order of magnitude lesser storage and activation memory. We further investigate the impact of using random and informed sampling techniques for dimensionality reduction in SEA.

In the early years of music information retrieval (MIR), research problems were often centered around conceptually simple tasks, and methods were evaluated on small, idealized data sets. A canonical example of this is genre recognition-i.e., Which one of n genres describes this song?-which was often evaluated on the GTZAN data set (1,000 musical excerpts balanced across ten genres) [1]. As task definitions were simple, so too were signal analysis pipelines, which often derived from methods for speech processing and recognition and typically consisted of simple methods for feature extraction, statistical modeling, and evaluation. When describing a research system, the expected level of detail was superficial: it was sufficient to state, e.g., the number of mel-frequency cepstral coefficients used, the statistical model (e.g., a Gaussian mixture model), the choice of data set, and the evaluation criteria, without stating the underlying software dependencies or implementation details. Because of an increased abundance of methods, the proliferation of software toolkits, the explosion of machine learning, and a focus shift toward more realistic problem settings, modern research systems are substantially more complex than their predecessors. Modern MIR researchers must pay careful attention to detail when processing metadata, implementing evaluation criteria, and disseminating results.

In the context of automatic speech recognition and acoustic event detection, an adaptive procedure named per-channel energy normalization (PCEN) has recently shown to outperform the pointwise logarithm of mel-frequency spectrogram (logmelspec) as an acoustic frontend. This letter investigates the adequacy of PCEN for spectrogram-based pattern recognition in far-field noisy recordings, both from theoretical and practical standpoints. First, we apply PCEN on various datasets of natural acoustic environments and find empirically that it Gaussianizes distributions of magnitudes while decorrelating frequency bands. Second, we describe the asymptotic regimes of each component in PCEN: temporal integration, gain control, and dynamic range compression. Third, we give practical advice for adapting PCEN parameters to the temporal properties of the noise to be mitigated, the signal to be enhanced, and the choice of time-frequency representation. As it converts a large class of real-world soundscapes into additive white Gaussian noise, PCEN is a computationally efficient frontend for robust detection and classification of acoustic events in heterogeneous environments.

The global COVID-19 pandemic motivates accelerated research to develop safe and efficacious vaccines. To address this need, we leveraged a biomaterial vaccine technology that consists of mesoporous silica rods (MSRs) that provide a sustained release of granulocyte-macrophage colony-stimulating factor (GM-CSF) and adjuvants to concentrate and mature antigen-presenting cells at the vaccine site. Here we explored the humoral responses resulting from the use of monophosphoryl lipid A (MPLA) as the adjuvant and SARS-CoV-2 spike proteins S1, S2, the nucleocapsid (N) protein, and receptor binding domain (RBD) as the target antigens. The dose of antigen and impact of pre-manufacturing of vaccines as versus loading antigen just-in-time was explored in these studies. Single shot MSR vaccines induced rapid and robust antibody titers to the presented antigens, even without the use of a boost, and sera from vaccinated animals demonstrated neutralizing activity against a SARS-CoV-2 pseudovirus. Overall, these results suggest the MSR vaccine system may provide potent protective immunity when utilized to present SARS-CoV-2 antigens.

Broad spectrum vaccines could provide a solution to the emergence of antibiotic resistant microbes, pandemics and engineered biothreat agents. Here, we describe a modular vaccine (composite infection vaccine technology (ciVAX)) which can be rapidly assembled and in which 4 of the 5 components are already approved for human use. ciVAX consists of an injectable biomaterial scaffold with factors to recruit and activate dendritic cells (DC) in vivo and microbeads conjugated with the broad-spectrum opsonin Fc-Mannose-binding Lectin (FcMBL) that is pre-bound to polysaccharide-rich cell wall antigens, such as the pathogen-associated molecular patterns (PAMPs) fractions, captured from whole inactivated bacteria. Vaccination of mice and rabbits with ciVAX generates potent humoral and T cell responses to PAMPs isolated from native antibiotic-resistant E. coli and S. aureus, and ciVAX protects mice and pigs against lethal E coli challenge in sepsis and septic shock models. In addition to the efficacy of ciVAX against homologous challenge, PAMPS isolated from an infected animal protects other animals against infection by heterologous challenge using different E. coli serotypes – demonstrating the potential for use of ciVAX in controlling pandemics. The advantage of the ciVAX technology is the strong immunogenicity with limited reactogenicity, the use of inactivated pathogens, and the modular manufacture using cGMP approved products which can be stockpiled ready for the next pandemic.One Sentence SummaryBiomaterial vaccine induces strong immunogenicity, weak reactogenicity, and protects from E. coli sepsis in rodents and pigs, and MRSA skin abscess.