Associations between SARS-CoV-2 Infection or COVID-19 Vaccination and Human Milk Composition: A Multi-Omics Approach (preprint)

The risk of contracting SARS-CoV-2 via human milk-feeding is virtually non-existent. Adverse effects of COVID-19 vaccination for lactating individuals are not different from the general population, and no evidence has been found that their infants exhibit adverse effects. Yet, there remains substantial hesitation among this population globally regarding the safety of these vaccines. Herein we aimed to determine if compositional changes in milk occur following infection or vaccination, including any evidence of vaccine components. Using an extensive multi-omics approach, we found that compared to unvaccinated individuals SARS-CoV-2 infection was associated with significant compositional differences in 67 proteins, 385 lipids, and 13 metabolites. In contrast, COVID-19 vaccination was not associated with any changes in lipids or metabolites, although it was associated with changes in 13 or fewer proteins. Compositional changes in milk differed by vaccine. Changes following vaccination were greatest after 1-6 hours for the mRNA-based Moderna vaccine (8 changed proteins), 3 days for the mRNA-based Pfizer (4 changed proteins), and adenovirus-based Johnson and Johnson (13 changed proteins) vaccines. Proteins that changed after both natural infection and Johnson and Johnson vaccine were associated mainly with systemic inflammatory responses. In addition, no vaccine components were detected in any milk sample. Together, our data provide evidence of only minimal changes in milk composition due to COVID-19 vaccination, with much greater changes after natural SARS-CoV-2 infection.

Precision medicine for pandemics: stratification of COVID-19 molecular phenotypes defined by topological analysis of global blood gene expression. (preprint)

Precision medicine offers a promising avenue for better therapeutic responses to pandemics such as COVID-19. This study leverages independent patient cohorts in Florence and Liege gathered under the umbrella of the DRAGON consortium for the stratification of molecular phenotypes associated with COVID-19 using topological analysis of global blood gene expression. Whole blood from 173 patients was collected and RNA was sequenced on the Novaseq platform. Molecular phenotypes were defined through topological analysis of gene expression relative to the biological network using the TopMD algorithm. The two cohorts from Florence and Liege allowed for independent validation of the findings in this study. Clustering of the topological maps of differential pathway activation revealed three distinct molecular phenotypes of COVID-19 in the Florence patient cohort, which were also observed in the Liege cohort. Cluster 1, was characterised by high activation of pathways associated with ESC pluripotency, NRF2, and TGF-B; receptor signalling. Cluster 2 displayed high activation of pathways including focal adhesion-PI3K-Akt-mTOR signalling and type I interferon induction and signalling, while Cluster 3 exhibited low IRF7-related pathway activation. TopMD was also used with the Drug-Gene Interaction Database (DGIdb), revealing pharmaceutical interventions targeting mechanisms across multiple phenotypes and individuals. The data illustrates the utility of molecular phenotyping from topological analysis of blood gene expression, and holds promise for informing personalised therapeutic strategies not only for COVID-19 but also for Disease X. Its potential transferability across multiple diseases highlights the value in pandemic response efforts, offering insights before large-scale clinical studies are initiated.

Understanding Human-COVID-19 Dynamics using Geospatial Big Data: A Systematic Literature Review (preprint)

The COVID-19 pandemic has changed human life. To mitigate the pandemic's impacts, different regions implemented various policies to contain COVID-19 and residents showed diverse responses. These human responses in turn shaped the uneven spatial-temporal spread of COVID-19. Consequently, the human-pandemic interaction is complex, dynamic, and interconnected. Delineating the reciprocal effects between human society and the pandemic is imperative for mitigating risks from future epidemics. Geospatial big data acquired through mobile applications and sensor networks have facilitated near-real-time tracking and assessment of human responses to the pandemic, enabling a surge in researching human-pandemic interactions. However, these investigations involve inconsistent data sources, human activity indicators, relationship detection models, and analysis methods, leading to a fragmented understanding of human-pandemic dynamics. To assess the current state of human-pandemic interactions research, we conducted a synthesis study based on 67 selected publications between March 2020 and January 2023. We extracted key information from each article across six categories, e.g., research area and time, data, methodological framework, and results and conclusions. Results reveal that regression models were predominant in relationship detection, featured in 67.16% of papers. Only two papers employed spatial-temporal models, notably underrepresented in the existing literature. Studies examining the effects of policies and human mobility on the pandemic's health impacts were the most prevalent, each comprising 12 articles (17.91%). Only 3 papers (4.48%) delved into bidirectional interactions between human responses and the COVID-19 spread. These findings shed light on the need for future research to spatially and temporally model the long-term, bidirectional causal relationships within human-pandemic systems.

Advancing Real-time Pandemic Forecasting Using Large Language Models: A COVID-19 Case Study (preprint)

Forecasting the short-term spread of an ongoing disease outbreak is a formidable challenge due to the complexity of contributing factors, some of which can be characterized through interlinked, multi-modality variables such as epidemiological time series data, viral biology, population demographics, and the intersection of public policy and human behavior. Existing forecasting model frameworks struggle with the multifaceted nature of relevant data and robust results translation, which hinders their performances and the provision of actionable insights for public health decision-makers. Our work introduces PandemicLLM, a novel framework with multi-modal Large Language Models (LLMs) that reformulates real-time forecasting of disease spread as a text reasoning problem, with the ability to incorporate real-time, complex, non-numerical information that previously unattainable in traditional forecasting models. This approach, through a unique AI-human cooperative prompt design and time series representation learning, encodes multi-modal data for LLMs. The model is applied to the COVID-19 pandemic, and trained to utilize textual public health policies, genomic surveillance, spatial, and epidemiological time series data, and is subsequently tested across all 50 states of the U.S. Empirically, PandemicLLM is shown to be a high-performing pandemic forecasting framework that effectively captures the impact of emerging variants and can provide timely and accurate predictions. The proposed PandemicLLM opens avenues for incorporating various pandemic-related data in heterogeneous formats and exhibits performance benefits over existing models. This study illuminates the potential of adapting LLMs and representation learning to enhance pandemic forecasting, illustrating how AI innovations can strengthen pandemic responses and crisis management in the future.

Discovery of orally bioavailable SARS-CoV-2 papain-like protease inhibitor as a potential treatment for COVID-19 (preprint)

The RNA-dependent RNA polymerase (RdRp), 3C-like protease (3CLpro), and papain-like protease (PLpro) are pivotal components in the viral life cycle of SARS-CoV-2, presenting as promising therapeutic targets. Currently, all FDA-approved antiviral drugs against SARS-CoV-2 are RdRp or 3CLpro inhibitors. However, the mutations causing drug resistance have been observed in RdRp and 3CLpro from SARS-CoV-2, which makes it necessary to develop antivirals with novel mechanisms. Through the application of a structure-based drug design (SBDD) approach, we discovered a series of novel potent non-covalent PLpro inhibitors with remarkable in vitro potency and in vivo PK properties. The co-crystal structures of PLpro with leads revealed that the residues E164 and Q269 around the S2 site are critical for improving the inhibitor\'s potency. The lead compound GZNL-P36 not only inhibited SARS-CoV-2 and its variants at the cellular level with EC50 ranging from 58.2 nM to 306.2 nM, but also inhibited HCoV-NL63 and HCoV-229E with EC50 of 81.6 nM and 2.66 M, respectively. Oral administration of the compound resulted in significantly improved survival and notable reductions in lung viral loads and lesions in SARS-CoV-2 infection mouse model, consistent with RNA-seq data analysis. Our results indicate that PLpro inhibitor is a promising SARS-CoV-2 therapy.

Loneliness during the last phase of the COVID-19 pandemic in China: Trajectories, predictors, and impact on adolescent mental health (preprint)

Background Little is known about the long-term courses of loneliness, associated risk factors and effect on mental health in adolescents during the COVID-19 pandemic. This study aimed to explore the trajectories of loneliness among Chinese adolescents during the last phase of the pandemic. We also aimed to identify risk factors in each loneliness course and the impact of loneliness on emotional problems, peer problems, hyperactivity and conduct problems. Methods  We conducted longitudinal analyses using four waves of data from 2347 Chinese adolescents covering a period of 20 months (October 2021 – May 2023). Loneliness was assessed using the UCLA 3-Item Loneliness Scale. The self-reported version of the Strengths and Difficulties Questionnaire was utilized to evaluate participants’ mental health outcomes. Growth mixture modelling was employed to identify latent classes of loneliness trajectories. Associated risk factors were investigated using multinomial logistic regression model. Mixed-effects logistic regression models were constructed to examine the long-term impact of loneliness classes on mental health outcomes. Results Three courses of loneliness were identified: Decreasing Low Loneliness (58.71%), Increasing Medium Loneliness (36.52%), and Increasing High Loneliness (4.77%). Risk factors for poorer loneliness trajectories included lack of physical exercise habits, poorer mental health literacy, medium or low perceived social support, having study difficulties, being female, higher grades, and lower economic status. Loneliness courses were associated with the severity and variability of emotional problems, peer problems, hyperactivity and conduct problems. Individuals in the higher loneliness classes experienced a significant increase in these mental health problems over time. Conclusions  During the last phase of the pandemic, a large proportion of adolescents in our study endured medium to high levels of loneliness with no signs of improvement. Both unfavorable loneliness trajectories adversely affected internalizing and externalizing problems and displayed an upward trend in these difficulties. Results highlight the importance of considering how to tackle loneliness both within the context of COVID-19 and more generally.

Event Detection from Social Media for Epidemic Prediction (preprint)

Social media is an easy-to-access platform providing timely updates about societal trends and events. Discussions regarding epidemic-related events such as infections, symptoms, and social interactions can be crucial for informing policymaking during epidemic outbreaks. In our work, we pioneer exploiting Event Detection (ED) for better preparedness and early warnings of any upcoming epidemic by developing a framework to extract and analyze epidemic-related events from social media posts. To this end, we curate an epidemic event ontology comprising seven disease-agnostic event types and construct a Twitter dataset SPEED with human-annotated events focused on the COVID-19 pandemic. Experimentation reveals how ED models trained on COVID-based SPEED can effectively detect epidemic events for three unseen epidemics of Monkeypox, Zika, and Dengue; while models trained on existing ED datasets fail miserably. Furthermore, we show that reporting sharp increases in the extracted events by our framework can provide warnings 4-9 weeks earlier than the WHO epidemic declaration for Monkeypox. This utility of our framework lays the foundations for better preparedness against emerging epidemics.

A Closer Look at Spatial-Slice Features Learning for COVID-19 Detection (preprint)

Conventional Computed Tomography (CT) imaging recognition faces two significant challenges: (1) There is often considerable variability in the resolution and size of each CT scan, necessitating strict requirements for the input size and adaptability of models. (2) CT-scan contains large number of out-of-distribution (OOD) slices. The crucial features may only be present in specific spatial regions and slices of the entire CT scan. How can we effectively figure out where these are located? To deal with this, we introduce an enhanced Spatial-Slice Feature Learning (SSFL++) framework specifically designed for CT scan. It aim to filter out a OOD data within whole CT scan, enabling our to select crucial spatial-slice for analysis by reducing 70% redundancy totally. Meanwhile, we proposed Kernel-Density-based slice Sampling (KDS) method to improve the stability when training and inference stage, therefore speeding up the rate of convergence and boosting performance. As a result, the experiments demonstrate the promising performance of our model using a simple EfficientNet-2D (E2D) model, even with only 1% of the training data. The efficacy of our approach has been validated on the COVID-19-CT-DB datasets provided by the DEF-AI-MIA workshop, in conjunction with CVPR 2024. Our source code will be made available.

Discovery of Natural Products to SARS-CoV-2 with Comparative Molecular Docking and Moderate Cell Proliferation Inhibition in vitro (preprint)

Based on screening in computational biology and biological in vitro assays, five natural products isolated from extracts of the herbal medicine toad skin, such as cinobufagin (CBFi), bufalin (BFi), arenobufagin (ABFi), telocinobufagin (TBFi), bufotalin (BFTi), were subjected to molecular docking calculations with the use of SARS-CoV-2 main protease (PDB 6LU7 and 7BTF) and top-scoring ligand-receptor complexes were obtained. The results showed that the binding energy of ABFi to the 3CL protein was -17.044kcal/mol, which was higher than CBFi and TBFi. However, the binding energy of ABFi to the RdRp protease was -23.250 kcal/mol, which was much lower than that of CBFi and TBFi, EVEN lower than that of ABFi to the 3CL protein. ABFi also has polar interactions with amino acids such as Glu811, Ser814, Ser681 and Thr680 of RdRp enzyme. The results revealed that ABFi had a moderate inhibitory effect on the cell proliferation of SARS-CoV-2 in vitro, with an inhibition rate of 61.12%, even weaker than Remdesivir. This new discovery provides us with new ideas for in-depth studies on the development of natural products with this class of structural generalizations as inhibitors of SARS-CoV-2, and provides an experimental basis for the next step of mechanistic studies.

CCQM-P199b: Interlaboratory comparability study of SARS-CoV-2 RNA copy number quantification (preprint)

Nucleic acid amplification tests including reverse transcription-quantitative PCR (RT-qPCR) are used to detect RNA from Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2), the causative agent of the Coronavirus disease 2019 (COVID-19) pandemic. Standardized measurements of RNA can facilitate comparable performance of laboratory tests in the absence of existing reference measurement systems early on in a pandemic. Interlaboratory study CCQM-P199b 'SARS-CoV-2 RNA copy number quantification' was designed to test the fitness-for-purpose of developed candidate reference measurement procedures (RMPs) for SARS-CoV-2 genomic targets in purified RNA materials, and was conducted under the auspices of the Consultative Committee for Amount of Substance: Metrology in Chemistry and Biology (CCQM) to evaluate the measurement comparability of national metrology institutes (NMIs) and designated institutes (DIs), thereby supporting international standardization. Twenty-one laboratories participated in CCQM-P199b and were requested to report the RNA copy number concentration, expressed in number of copies per microliter, of the SARS-CoV-2 nucleocapsid (N) gene partial region (NC_045512.2: 28274-29239) and envelope (E) gene (NC_045512.2: 26245-26472) (optional measurement) in samples consisting of in vitro transcribed RNA or purified RNA from lentiviral constructs. Materials were provided in two categories: lower concentration (approximately 10 x 1 - 10 x 4/uL in aqueous solution containing human RNA background) and high concentration (approximately 10 x 9/uL in aqueous solution without any other RNA background). For the measurement of N gene concentration in the lower concentration study materials, the majority of laboratories (n = 17) used one-step reverse transcription-digital PCR (RT-dPCR), with three laboratories applying two-step RT-dPCR and one laboratory RT-qPCR. Sixteen laboratories submitted results for E gene concentration. Reproducibility (% CV or equivalent) for RT-dPCR ranged from 19 % to 31 %. Measurements of the high concentration study material by orthogonal methods (isotope dilution-mass spectrometry and single molecule flow cytometry) and a gravimetrically linked lower concentration material were in a good agreement, suggesting a lack of overall bias in RT-dPCR measurements. However methodological factors such as primer and probe (assay) sequences, RT-dPCR reagents and dPCR partition volume were found to be potential sources of interlaboratory variation which need to be controlled when applying this technique. This study demonstrates that the accuracy of RT-dPCR is fit-for-purpose as a RMP for viral RNA target quantification in purified RNA materials and highlights where metrological approaches such as the use of in vitro transcribed controls, orthogonal methods and measurement uncertainty evaluation can support standardization of molecular methods.

Identifying immune signatures of common exposures through co-occurrence of T-cell receptors in tens of thousands of donors (preprint)

Memory T cells are records of clonal expansion from prior immune exposures, such as infections, vaccines and chronic diseases like cancer. A subset of the receptors of these expanded T cells in a typical immune repertoire are highly public, i.e., present in many individuals exposed to the same exposure. For the most part, the exposures associated with these public T cells are unknown. To identify public T-cell receptor signatures of immune exposures, we mined the immunosequencing repertoires of tens of thousands of donors to define clusters of co-occurring T cells. We first built co-occurrence clusters of T cells responding to antigens presented by the same Human Leukocyte Antigen (HLA) and then combined those clusters across HLAs. Each cross-HLA cluster putatively represents the public T-cell signature of a single prevalent exposure. Using repertoires from donors with known serological status for 7 prevalent exposures (HSV-1, HSV-2, EBV, Parvovirus, Toxoplasma gondii, Cytomegalovirus and SARS CoV-2), we identified a single T-cell cluster strongly associated with each exposure and used it to construct a highly sensitive and specific diagnostic model for the exposure. These T-cell clusters constitute the public immune responses to prevalent exposures, 7 known and many others unknown. By learning the exposure associations for more T cell clusters, this approach could be used to derive a ledger of a person's past and present immune exposures.

Multivalent S2-subunit Vaccines Provide Broad Protection Against Clade 1 Sarbecoviruses (preprint)

The continuing emergence of immune evasive SARS-CoV-2 variants and the previous SARS-CoV-1 outbreak have accentuated the need for broadly protective sarbecovirus vaccines. Targeting the conserved S2-subunit of SARS-CoV-2 is a particularly promising approach to elicit broad protection. Here, expanding on our previous work with S2-based vaccines, we developed a nanoparticle vaccine displaying multiple copies of the SARS-CoV-1 S2 subunit. This vaccine alone, or as a cocktail with a SARS-CoV-2 S2 subunit vaccine, protected transgenic K18-hACE2 mice from challenges with Omicron subvariant XBB as well as several sarbecoviruses identified as having pandemic potential including the bat sarbecovirus WIV1, BANAL-236, and a pangolin sarbecovirus. Challenge studies in Fc-g receptor knockout mice revealed that antibody-based cellular effector mechanisms played a role in protection elicited by these vaccines. These results demonstrate that our S2-based vaccines provide broad protection against clade 1 sarbecoviruses and offer insight into the mechanistic basis for protection.

Evidence-Driven Retrieval Augmented Response Generation for Online Misinformation (preprint)

The proliferation of online misinformation has posed significant threats to public interest. While numerous online users actively participate in the combat against misinformation, many of such responses can be characterized by the lack of politeness and supporting facts. As a solution, text generation approaches are proposed to automatically produce counter-misinformation responses. Nevertheless, existing methods are often trained end-to-end without leveraging external knowledge, resulting in subpar text quality and excessively repetitive responses. In this paper, we propose retrieval augmented response generation for online misinformation (RARG), which collects supporting evidence from scientific sources and generates counter-misinformation responses based on the evidences. In particular, our RARG consists of two stages: (1) evidence collection, where we design a retrieval pipeline to retrieve and rerank evidence documents using a database comprising over 1M academic articles; (2) response generation, in which we align large language models (LLMs) to generate evidence-based responses via reinforcement learning from human feedback (RLHF). We propose a reward function to maximize the utilization of the retrieved evidence while maintaining the quality of the generated text, which yields polite and factual responses that clearly refutes misinformation. To demonstrate the effectiveness of our method, we study the case of COVID-19 and perform extensive experiments with both in- and cross-domain datasets, where RARG consistently outperforms baselines by generating high-quality counter-misinformation responses.

SARS-CoV-2 infection activates inflammatory macrophages in vascular immune organoids (preprint)

SARS-CoV-2 provokes devastating tissue damage by cytokine release syndrome and leads to multi-organ failure. Modeling the process of immune cell activation and subsequent tissue damage is a significant task. Organoids from human tissues advanced our understanding of SARS-CoV-2 infection mechanisms though, they are missing crucial components: immune cells and endothelial cells. This study aims to generate organoids with these components. We established vascular immune organoids from human pluripotent stem cells and examined the effect of SARS-CoV-2 infection. We demonstrated that infections activated inflammatory macrophages. Notably, the upregulation of interferon signaling supports macrophages role in cytokine release syndrome. We propose vascular immune organoids are a useful platform to model and discover factors that ameliorate SARS-CoV-2-mediated cytokine release syndrome.

Two fitness inference schemes compared using allele frequencies from 1,068,391 sequences sampled in the UK during the COVID-19 pandemic (preprint)

Throughout the course of the SARS-CoV-2 pandemic, genetic variation has contributed to the spread and persistence of the virus. For example, various mutations have allowed SARS-CoV-2 to escape antibody neutralization or to bind more strongly to the receptors that it uses to enter human cells. Here, we compared two methods that estimate the fitness effects of viral mutations using the abundant sequence data gathered over the course of the pandemic. Both approaches are grounded in population genetics theory but with different assumptions. One approach, tQLE, features an epistatic fitness landscape and assumes that alleles are nearly in linkage equilibrium. Another approach, MPL, assumes a simple, additive fitness landscape, but allows for any level of correlation between alleles. We characterized differences in the distributions of fitness values inferred by each approach and in the ranks of fitness values that they assign to sequences across time. We find that in a large fraction of weeks the two methods are in good agreement as to their top-ranked sequences, i.e., as to which sequences observed that week are most fit. We also find that agreement between ranking of sequences varies with genetic unimodality in the population in a given week.

Navigating stroke care during the COVID-19 pandemic: A scoping review (preprint)

Background: Previous viral outbreaks have highlighted implications for the management of complex health conditions. This study delves into the repercussions of the COVID-19 pandemic on stroke care, by examining evidence of shifts in healthcare utilization, the enduring effects on post-stroke recovery, and the overall quality of life experienced by stroke survivors. Methods: A scoping review was conducted following the Joanna Briggs Institute Methodology for Scoping Reviews. The search strategy encompassed electronic databases (APA PsycInfo, Embase, Medline, and CINAHL). English language articles published between December 2019 and January 2022 were included, focusing on individuals who experienced a stroke during the COVID-19 pandemic. Data extraction involved identifying study characteristics and significant findings, facilitating a qualitative and narrative synthesis of the gathered evidence. Results: Seven domain summaries were identified. They all described the aspects of systemic transformations in stroke care during the COVID-19 pandemic: (1) patient behavior and awareness; (2) telemedicine and remote care; (3) delays in treatment; (4) impact on healthcare resources; (5) quality of care; (6) changes in stroke severity; and (7) reduction in stroke admissions. Conclusions: This study underscored the critical need to encourage swift patient response to acute stroke symptoms, by finding new avenues for treatment, mitigating hospital-related infection fears, and advocating for the establishment of centralized stroke centers. These measures are integral to optimizing stroke care delivery and ensuring timely interventions, particularly in the challenging context of a pandemic.

Differentiating Immune Checkpoint Inhibitor-Related Pneumonitis From COVID-19 Pneumonia Using a CT-based Radiomics Nomogram (preprint)

Objectives This study aimed to develop and validate a radiomics nomogram that effectively distinguishes between immune checkpoint inhibitor-related pneumonitis (CIP) and COVID-19 pneumonia using radiographic imaging features. Methods We included 97 patients in this study, identifying 269 pneumonia lesions—159 from COVID-19 and 110 from CIP. The dataset was randomly divided into a training set (70% of the data) and a validation set (30%). We extracted radiomics features from corticomedullary and nephrographic phase-contrast computed tomography (CT) images, constructed a radiomics signature, and calculated a radiomics score (Rad-score). Using these features, we built models with three classifiers and assessed demographics and CT findings to create a clinical factors model. We then constructed a radiomics nomogram that combines the Rad-score with independent clinical factors and evaluated its performance in terms of calibration, discrimination, and clinical usefulness. Results In constructing the radiomics signature, 33 features were critical for differentiating between CIP and COVID-19 pneumonia. The support vector machine classifier was the most accurate of the three classifiers used. The Rad-score, gender, lesion location, radiological features, and lesion borders were included in the nomogram. The nomogram demonstrated superior predictive performance, significantly outperforming the clinical factors model in the training set (AUC comparison, p = 0.02638). Calibration curves indicated good fit in both training and validation sets, and the nomogram displayed greater net benefit compared to the clinical model. Conclusion The radiomics nomogram emerges as a noninvasive, quantitative tool with significant potential to differentiate between CIP and COVID-19 pneumonia. It enhances diagnostic accuracy and supports radiologists, especially in overburdened medical systems, through the use of machine learning predictions.