Search

Abstract: Recently, combinations of generative and Bayesian deep learning have been introduced in particle physics for both fast detector simulation and inference tasks. These neural networks aim to quantify the uncertainty on the generated distribution originating from limited training statistics. The interpretation of a distribution-wide uncertainty however remains ill-defined. We show a clear scheme for quantifying the calibration of Bayesian generative machine learning models. For a Continuous Normalizing Flow applied to a low-dimensional toy example, we evaluate the calibration of Bayesian uncertainties from either a mean-field Gaussian weight posterior, or Monte Carlo sampling network weights, to gauge their behaviour on unsteady distribution edges. Well calibrated uncertainties can then be used to roughly estimate the number of uncorrelated truth samples that are equivalent to the generated sample and clearly indicate data amplification for smooth features of the distribution.

Machine learning-based unfolding has enabled unbinned and high-dimensional differential cross section measurements. Two main approaches have emerged in this research area; one based on discriminative models and one based on generative models. The main advantage of discriminative models is that they learn a small correction to a starting simulation while generative models scale better to regions of phase space with little data. We propose to use Schrödinger bridges and diffusion models to create sbunfold, an unfolding approach that combines the strengths of both discriminative and generative models. The key feature of sbunfold is that its generative model maps one set of events into another without having to go through a known probability density as is the case for normalizing flows and standard diffusion models. We show that sbunfold achieves excellent performance compared to state of the art methods on a synthetic Z+jets dataset.

Abstract Targeted (nano-)drug delivery is essential for treating respiratory diseases, which are often confined to distinct lung regions. However, spatio-temporal profiling of drugs or nanoparticles (NPs) and their interactions with lung macrophages remains unresolved. Here, we present LungVis 1.0, an AI-powered imaging ecosystem that integrates light sheet fluorescence microscopy with deep learning-based image analysis pipelines to map NP deposition and dosage holistically and quantitatively across bronchial and alveolar (acinar) regions in murine lungs for widely-used bulk-liquid and aerosol-based delivery methods. We demonstrate that bulk-liquid delivery results in patchy NP distribution with elevated bronchial doses, whereas aerosols achieve uniform deposition reaching distal alveoli. Furthermore, we reveal that lung tissue-resident macrophages (TRMs) are dynamic, actively patrolling and redistributing NPs within alveoli, contesting the conventional paradigm of TRMs as static entities. LungVis 1.0 provides an advanced framework for exploring pulmonary delivery dynamics and deepening insights into TRM-mediated lung immunity.

A critical question concerning generative networks applied to event generation in particle physics is if the generated events add statistical precision beyond the training sample. We show for a simple example with increasing dimensionality how generative networks indeed amplify the training statistics. We quantify their impact through an amplification factor or equivalent numbers of sampled events.

Summary: Safe, effective, and low-cost oral antiviral therapies are needed to treat those at high risk for developing severe COVID-19. To that end, we performed a high-throughput screen to identify non-peptidic, non-covalent inhibitors of the SARS-CoV-2 main protease (Mpro), an essential enzyme in viral replication. NZ-804 was developed from a screening hit through iterative rounds of structure-guided medicinal chemistry. NZ-804 potently inhibits SARS-CoV-2 Mpro (0.009 μM IC50) as well as SARS-CoV-2 replication in human lung cell lines (0.008 μM EC50) and primary human airway epithelial cell cultures. Antiviral activity is maintained against distantly related sarbecoviruses and endemic human CoV OC43. In SARS-CoV-2 mouse and hamster disease models, NZ-804 therapy given once or twice daily significantly diminished SARS-CoV-2 replication and pathogenesis. NZ-804 synthesis is low cost and uncomplicated, simplifying global production and access. These data support the exploration of NZ-804 as a therapy for COVID-19 and future emerging sarbecovirus infections.

Abstract: Motivated by the high computational costs of classical simulations, machine-learned generative models can be extremely useful in particle physics and elsewhere. They become especially attractive when surrogate models can efficiently learn the underlying distribution, such that a generated sample outperforms a training sample of limited size. This kind of GANplification has been observed for simple Gaussian models. We show the same effect for a physics simulation, specifically photon showers in anelectromagnetic calorimeter.

During the COVID-19 pandemic, German early childhood education and care (ECEC) institutions had to limit their provision of ECEC, implement protective measures, and handle new organizational tasks. Data from two longitudinal surveys (October 2020 to July 2021) among ECEC managers and pedagogical staff were analysed. Limited staff resources, limited access, and high pandemic-related challenges (i.e. difficulties and conflicts related to implementing protective measures), were negatively associated with the frequency of pedagogical practices (Models 1, 2). Manager ratings suggested increased developmental needs for children with low socio-economic status; also higher frequency of pedagogical practices at early stages of the pandemic (T1) was associated with lower increase of developmental needs at a later stage (T2, about 5 months later). In sum, this contribution provides evidence about how the COVID-19 pandemic might have affected quality in ECEC provision on both the structural and interaction level and how this, subsequently, impacted child outcomes.

Abstract The two p53 homologues p63 and p73 regulate transcriptional programs in epithelial tissues and several cell types in these tissues express both proteins. All members of the p53 family form tetramers in their active state through a dedicated oligomerization domain that structurally assembles as a dimer of dimers. The oligomerization domain of p63 and p73 share a high sequence identity, but the p53 oligomerization domain is more divergent and it lacks a functionally important C-terminal helix present in the other two family members. Based on these structural differences, p53 does not hetero-oligomerize with p63 or p73. In contrast, p63 and p73 form hetero-oligomers of all possible stoichiometries, with the hetero-tetramer built from a p63 dimer and a p73 dimer being thermodynamically more stable than the two homo-tetramers. This predicts that in cells expressing both proteins a p632/p732 hetero-tetramer is formed. So far, the tools to investigate the biological function of this hetero-tetramer have been missing. Here we report the generation and characterization of Designed Ankyrin Repeat Proteins (DARPins) that bind with high affinity and selectivity to the p632/p732 hetero-tetramer. Using these DARPins we were able to confirm experimentally the existence of this hetero-tetramer in epithelial mouse and human tissues and show that its level increases in squamous cell carcinoma.

The genetic diversity of the coronavirus (CoV) family poses a significant challenge for drug discovery and development. Traditional antiviral drugs often target specific viral proteins from specific viruses which limits their use, especially against novel emerging viruses. Antivirals with broad-spectrum activity overcome this limitation by targeting highly conserved regions or catalytic domains within viral proteins that are essential for replication. For rapid identification of small molecules with broad antiviral activity, assays with viruses representing family-wide genetic diversity are needed. Viruses engineered to express a reporter gene (i.e. luminescence, fluorescence, etc.) can increase the efficiency, sensitivity or precision of drug screening over classical measures of replication like observation of cytopathic effect or measurement of infectious titers. We have previously developed reporter virus systems for multiple other endemic, pandemic, epidemic and enzootic CoV. Human CoV OC43 (HCoV-OC43) is a human endemic CoV that causes respiratory infection with age-related exacerbations of pathogenesis. Here, we describe the development of a novel recombinant HCoV-OC43 reporter virus that expresses nano-luciferase (HCoV-OC43 nLuc), and its potential application for screening of antivirals against CoV.

A critical question concerning generative networks applied to event generation in particle physics is if the generated events add statistical precision beyond the training sample. We show for a simple example with increasing dimensionality how generative networks indeed amplify the training statistics. We quantify their impact through an amplification factor or equivalent numbers of sampled events.

This meta-analysis examines the methodological tendencies of scientific research about political communication on Twitter published in journals based on Spanish-speaking countries between 2019 and 2021. It covers all the journals indexed in the JCR and the two first quartiles of Scopus index, producing a universe of 1.233 articles, of which 51 addressed political communication on Twitter. The focus is set on the following methodological parameters: sample types and construction, time periods, geographical areas, methodologies, multimedia analysis, general research themes and approaches, and software resources. The aim is to provide a map of recent research in the Spanish-speaking scientific community and to identify widespread methodological trends. A significant trend of software dependency was identified, i.e. published research often consists of the mere application of a certain software to a certain dataset, without enough discussion on the possible biases introduced by the software and the sampling methods.

Significant advances in deep learning have led to more widely used and precise neural network-based generative models such as Generative Adversarial Networks (Gans). We introduce a post-hoc correction to deep generative models to further improve their fidelity, based on the Deep neural networks using the Classification for Tuning and Reweighting (Dctr) protocol. The correction takes the form of a reweighting function that can be applied to generated examples when making predictions from the simulation. We illustrate this approach using Gans trained on standard multimodal probability densities as well as calorimeter simulations from high energy physics. We show that the weighted Gan examples significantly improve the accuracy of the generated samples without a large loss in statistical power. This approach could be applied to any generative model and is a promising refinement method for high energy physics applications and beyond.

Background: Assessment of age-associated disorders has become increasingly important. Methods: In a clinical setting, people with intellectual disability with and without dementia were assessed retrospectively using the Neuropsychological Test Battery (NTB) and the Dementia Questionnaire for People with Learning Disabilities (DLD) at two different times to analyse neuropsychological changes and diagnostic validity. One group (n = 44) was assessed with both instruments, while the DLD was applied in 71 patients. Results: In the NTB (n = 44), only patients with dementia (n = 26) showed a decline in the NTB total score and three subscales. Receiver operating characteristic analysis revealed a diagnostic sensitivity of 0.67, a specificity of 0.81, and an area under the curve (AUC) of 0.767. In the DLD group (n = 71), only those with dementia displayed a decrease in the cognitive and social scale; diagnostic sensitivity and specificity values were low (0.61/0.63) and the AUC was 0.704. Conclusions: Neuropsychological assessment was sensitive to detect cognitive changes over time. Sensitivity values of both instruments suggest a reassessment at a later time point.

Este meta-estudio analiza las tendencias metodológicas de los artículos científicos centrados en la comunicación política en Twitter publicados en español entre 2019 y 2021 en las revistas de mayor índice de impacto (JCR y Scopus Q1 y Q2). De un universo de 1233 artículos, se analiza una muestra de 51 publicados en siete revistas. Se presta atención a los siguientes parámetros metodológicos: tipos y construcciones muestrales, períodos temporales, ámbitos geográficos estudiados, metodologías de análisis empleadas, análisis multimediales, temas y planteamientos generales de investigación, y empleo de software. El objetivo es ofrecer un mapa de la investigación científica reciente e identificar carencias o defectos metodológicos extendidos. De entre las conclusiones destaca una incidencia importante de artículos que manifiestan software dependency, es decir, que consisten principalmente en la mera aplicación de un programa o aplicación informática a un conjunto de datos.

Lei et al. systematically characterized the epistasis among natural mutations in the neuraminidase of human influenza H3N2 virus, which provide insights into the biophysical constraints that shaped its evolution trajectory over the past half-century.

Recently, combinations of generative and Bayesian deep learning have been introduced in particle physics for both fast detector simulation and inference tasks. These neural networks aim to quantify the uncertainty on the generated distribution originating from limited training statistics. The interpretation of a distribution-wide uncertainty however remains ill-defined. We show a clear scheme for quantifying the calibration of Bayesian generative machine learning models. For a Continuous Normalizing Flow applied to a low-dimensional toy example, we evaluate the calibration of Bayesian uncertainties from either a mean-field Gaussian weight posterior, or Monte Carlo sampling network weights, to gauge their behaviour on unsteady distribution edges. Well calibrated uncertainties can then be used to roughly estimate the number of uncorrelated truth samples that are equivalent to the generated sample and clearly indicate data amplification for smooth features of the distribution.

Abstract Molecular understanding of osteogenic differentiation (OD) of human bone marrow-derived mesenchymal stem cells (hBMSCs) is important for regenerative medicine and has direct implications for cancer. We report that the RNF4 ubiquitin ligase is essential for OD of hBMSCs, and that RNF4-deficient hBMSCs remain as stalled progenitors. Remarkably, incubation of RNF4-deficient hBMSCs in conditioned media of differentiating hBMSCs restored OD. Transcriptional analysis of RNF4-dependent gene signatures identified two secreted factors that act downstream of RNF4 promoting OD: (1) BMP6 and (2) the BMP6 co-receptor, RGMb (Dragon). Indeed, knockdown of either RGMb or BMP6 in hBMSCs halted OD, while only the combined co-addition of purified RGMb and BMP6 proteins to RNF4-deficient hBMSCs fully restored OD. Moreover, we found that the RNF4-RGMb-BMP6 axis is essential for survival and tumorigenicity of osteosarcoma and therapy-resistant melanoma cells. Importantly, patient-derived sarcomas such as osteosarcoma, Ewing sarcoma, liposarcomas, and leiomyosarcomas exhibit high levels of RNF4 and BMP6, which are associated with reduced patient survival. Overall, we discovered that the RNF4~BMP6~RGMb axis is required for both OD and tumorigenesis.

Oncogenic transformation of lung epithelial cells is a multistep process, frequently starting with the inactivation of tumour suppressors and subsequent development of activating mutations in proto‐oncogenes, such as members of the PI3K or MAPK families. Cells undergoing transformation have to adjust to changes, including altered metabolic requirements. This is achieved, in part, by modulating the protein abundance of transcription factors. Here, we report that the ubiquitin carboxyl‐terminal hydrolase 28 (USP28) enables oncogenic reprogramming by regulating the protein abundance of proto‐oncogenes such as c‐JUN, c‐MYC, NOTCH and ∆NP63 at early stages of malignant transformation. USP28 levels are increased in cancer compared with in normal cells due to a feed‐forward loop, driven by increased amounts of oncogenic transcription factors such as c‐MYC and c‐JUN. Irrespective of oncogenic driver, interference with USP28 abundance or activity suppresses growth and survival of transformed lung cells. Furthermore, inhibition of USP28 via a small‐molecule inhibitor resets the proteome of transformed cells towards a ‘premalignant’ state, and its inhibition synergizes with clinically established compounds used to target EGFRL858R‐, BRAFV600E‐ or PI3KH1047R‐driven tumour cells. Targeting USP28 protein abundance at an early stage via inhibition of its activity is therefore a feasible strategy for the treatment of early‐stage lung tumours, and the observed synergism with current standard‐of‐care inhibitors holds the potential for improved targeting of established tumours.

Ribosomal biogenesis and protein synthesis are deregulated in most cancers, suggesting that interfering with translation machinery may hold significant therapeutic potential. Here, we show that loss of the tumor suppressor adenomatous polyposis coli (APC), which constitutes the initiating event in the adenoma carcinoma sequence for colorectal cancer (CRC), induces the expression of RNA polymerase I (RNAPOL1) transcription machinery, and subsequently upregulates ribosomal DNA (rDNA) transcription. Targeting RNAPOL1 with a specific inhibitor, CX5461, disrupts nucleolar integrity, and induces a disbalance of ribosomal proteins. Surprisingly, CX5461‐induced growth arrest is irreversible and exhibits features of senescence and terminal differentiation. Mechanistically, CX5461 promotes differentiation in an MYC‐interacting zinc‐finger protein 1 (MIZ1)‐ and retinoblastoma protein (Rb)‐dependent manner. In addition, the inhibition of RNAPOL1 renders CRC cells vulnerable towards senolytic agents. We validated this therapeutic effect of CX5461 in murine‐ and patient‐derived organoids, and in a xenograft mouse model. These results show that targeting ribosomal biogenesis together with targeting the consecutive, senescent phenotype using approved drugs is a new therapeutic approach, which can rapidly be transferred from bench to bedside.