No abstract

In the absence of nationwide mass testing for an emerging health crisis, alternative approaches could provide necessary information efficiently to aid policy makers and health bodies when dealing with a pandemic. The following work presents a methodology by which Twitter data surrounding the first wave of the COVID-19 pandemic in the UK is harvested and analysed using two main approaches. The first is an investigation into localized outbreak predictions by developing a prototype early-warning system using the distribution of total tweet volume. The temporal lag between the rises in the number of COVID-19 related tweets and officially reported deaths by Public Health England (PHE) is observed to be 6–27 days for various UK cities which matches the temporal lag values found in the literature. To better understand the topics of discussion and attitudes of people surrounding the pandemic, the second approach is an in-depth behavioural analysis assessing the public opinion and response to government policies such as the introduction of face-coverings. Using topic modelling, nine distinct topics are identified within the corpus of COVID-19 tweets, of which the themes ranged from retail to government bodies. Sentiment analysis on a subset of mask related tweets revealed sentiment spikes corresponding to major news and announcements. A Named Entity Recognition (NER) algorithm is trained and applied in a semi-supervised manner to recognise tweets containing location keywords within the unlabelled corpus and achieved a precision of 81.6%. Overall, these approaches allowed extraction of temporal trends relating to PHE case numbers, popular locations in relation to the use of face-coverings, and attitudes towards face-coverings, vaccines and the national ‘Test and Trace’ scheme.

In the study of grain-surface chemistry in the interstellar medium, there exists much uncertainty regarding the reaction mechanisms with few constraints on the abundances of grain-surface molecules. Bayesian inference can be performed to determine the likely reaction rates. In this work, we consider methods for reducing the computational expense of performing Bayesian inference on a reaction network by looking at the geometry of the network. Two methods of exploiting the topology of the reaction network are presented. One involves reducing a reaction network to just the reaction chains with constraints on them. After this, new constraints are added to the reaction network and it is shown that one can separate this new reaction network into subnetworks. The fact that networks can be separated into subnetworks is particularly important for the reaction networks of interstellar complexorganic molecules, whose surface reaction networks may have hundreds of reactions. Both methods allow the maximum-posterior reaction rate to be recovered with minimal bias.

There exists much uncertainty surrounding interstellar grain-surface chemistry. One of the major reaction mechanisms is grain-surface diffusion for which the binding energy parameter for each species needs to be known. However, these values vary significantly across the literature which can lead to debate as to whether or not a particular reaction takes place via diffusion. In this work we employ Bayesian inference to use available ice abundances to estimate the reaction rates of the reactions in a chemical network that produces glycine. Using this we estimate the binding energy of a variety of important species in the network, by assuming that the reactions take place via diffusion. We use our understanding of the diffusion mechanism to reduce the dimensionality of the inference problem from 49 to 14, by demonstrating that reactions can be separated into classes. This dimensionality reduction makes the problem computationally feasible. A neural network statistical emulator is used to also help accelerate the Bayesian inference process substantially. The binding energies of most of the diffusive species of interest are found to match some of the disparate literature values, with the exceptions of atomic and diatomic hydrogen. The discrepancies between these two species are related to the limitations of the physical and chemical models. However, the use of a dummy reaction of the form H + X ⟶ HX is found to somewhat reduce the discrepancy with the binding energy of atomic hydrogen. Using the inferred binding energies in the full gas–grain version of UCLCHEM results in almost all the molecular abundances being recovered.

Tuberculous granulomas are highly dynamic structures reflecting the complex host–mycobacterium interactions. The objective of this study was to compare granuloma development at the site of vaccination with BCG and its recombinant derivatives in goats. To characterize the host response, epithelioid cells, multinucleated giant cells (MNGC), T cell subsets, B cells, plasma cells, dendritic cells and mycobacterial antigen were labelled by immunohistochemistry, and lipids and acid-fast bacteria (AFB) were labelled by specific staining. Granulomas with central caseous necrosis developed at the injection site of most goats though lesion size and extent of necrosis differed between vaccine strains. CD4+ T and B cells were more scarce and CD8+ cells were more numerous in granulomas induced by recombinant derivatives compared to their parental BCG strain. Further, the numbers of MNGCs and cells with lipid bodies were markedly lower in groups administered with recombinant BCG strains. Microscopic detection of AFB and mycobacterial antigen was rather frequent in the area of central necrosis, however, the isolation of bacteria in culture was rarely successful. In summary, BCG and its recombinant derivatives induced reproducibly subcutaneous caseous granulomas in goats that can be easily monitored and surgically removed for further studies. The granulomas reflected the genetic modifications of the recombinant BCG-derivatives and are therefore suitable models to compare reactions to different mycobacteria or TB vaccines.