Campylobacter is currently the most common cause of acute bacterial diarrhoea in Australia among all the notified enteric pathogens with more than 15,000 cases each year. The incidence of notified campylobacteriosis has steadily increased during the past 15 years from 67.0/100,000 population in 1991 to 121.4/100,000 in 2005, though the factors contributing to this increase had not been studied. Adjusting for under-reporting there are, at this point in time, an estimated 225,000 infections occurring each year in Australia, most of which are sporadic in nature. Much of our knowledge in Australia about risk factors for sporadic disease has been based on overseas literature. Prior to the studies undertaken in this thesis, the epidemiology of Campylobacter infection had not been thoroughly studied in Australia, nor had there been any national studies examining risk factors for locally-acquired infection. The broad aim of this thesis was to examine in depth the descriptive epidemiology of Campylobacter infection in Australia, explore the reasons for the sustained increase in incidence of infection and to identify the major risk factors for locally acquired infection using a multi-centre case-control study design. The descriptive study of the epidemiology of campylobacteriosis in Australia was based on Australian notifiable disease surveillance data collected over a 15-year period between 1991 and 2005. This study described the key epidemiological characteristics of this disease in Australia and identified some significant differences in incidence trends across states and territories and among different age groups which had not been previously reported. The study identified gaps in our knowledge of this disease in Australia and made recommendations for future research including the investigation of factors associated with the decline in incidence of infection among children aged  4 years and further studies to identify age and sex-specific risk factors for infection. The issue of seasonality, transmission routes and infection was addressed and areas for further research were specified including longitudinal studies at a regional level that incorporate a comparison of human, animal and environmental genotypes. This study also provided strong compelling evidence to support the hypothesis that the increase in notification rates in Australia during this period represented a real increase in the incidence of infection and that the main driving force behind this rise has been the ongoing increase in chicken consumption among the Australian public. The multi-centre case-control study, involving 1,714 participants  5 years of age, identified the major foodborne and non-foodborne risk factors for Campylobacter infection among the general population in Australia. This study confirmed that chicken meat is a major source of sporadic infection in this country and is responsible for almost one-third of all cases that occur in the Australian community each year. Other independent risk factors for sporadic infection in Australia included consumption of offal and ownership of domestic dogs or chickens aged  6 months. The Nagelkerke R² value of 16% for the final multivariable model indicated a considerable proportion of our case-patients had unexplained risk factors. The combined population attributable risk (PAR) estimate for the independent foodborne risk factors in this study was 31%, which is considerably less than the 75% to 80% of cases in the general population which are thought to be caused through foodborne transmission. Possible explanations for these results include the likelihood that a proportion of foodborne transmission in Australia occurs through food vehicles other than chicken due to cross-contamination from raw products, and the likelihood that much of the population attributable risk that is unaccountered for, may in fact be due to inherent limitations of study design resulting in systematic errors (information bias) and possibly reduced estimates of effect. The burden of illness among the general population in Australia attributable to different independent risk factors was estimated using a novel method developed specifically for this study. Briefly, community incidence data was coupled with PAR data from the case-control study and simulation techniques were used to: (i) estimate the number of infections attributable to specific risk factors, and (ii) derive credible intervals for these estimates by modeling the uncertainty in each variable component. This model of using case-control data in conjunction with pre-existing surveillance data provides researchers with a simple but robust tool for conducting source attribution studies on enteric pathogens. In conclusion, the studies undertaken in this thesis have made important contributions to our understanding of the epidemiology of sporadic Campylobacter infection in Australia.

The use of Positive Matrix Factorization (PMF) in pollution source apportionment (PSA) is examined and illustrated. A study of its settings is conducted in order to optimize them in the context of PSA. The use of a priori information in PMF is examined, in the form of target factor profiles and pulling profile elements to zero. A Bayesian model using lognormal prior distributions for source profiles and source contributions is fit and examined.

The impacts of transported and locally-produced pollutants on western US air quality during summer 2008 are studied using the multi-scale Sulfur Transport and Deposition Modeling system. Transported background (TBG) is an indicator of the influences from extra-regional emissions or the lower stratosphere. The magnitude of TBG is expected to increase as the emissions from international sources grow. This trend is especially important in the context of US air quality standards, which tend to become more stringent to protect human health and ecosystems. Forward sensitivity simulations in which the model boundary conditions and emissions are perturbed show that TBG strongly and extensively affect western US surface ozone (more than half of the total), compared to other contributors to background ozone (North American, NA, biomass burning, BB and biogenic emissions), and the impacts differ among various geographical regions and land types. The stratospheric ozone impacts are weak. The TBG ozone contributes most to western US ozone among all TBG species, and TBG peroxyacetyl nitrate is the most important species among ozone precursors. Compared to monthly mean 8-hour daily maximum ozone, the secondary standard metric "W126 monthly index" shows larger responses to TBG perturbations and stronger non-linearity to the size of perturbations. Overall the model-estimated TBG impacts negatively correlate to the vertical resolution and positively correlate to the horizontal resolution. The estimated TBG impacts weakly depend on the magnitude of uncertainties in the US anthropogenic emissions. The transport/subsidence processes that link airmasses aloft with the surface pollution level are analyzed by trajectories, time-lag correlation and adjoint sensitivity analyses. Various types of observations are used to identify source regions and transport processes, and to improve model prediction using the four-dimensional variational data assimilation during a long-range transport episode. The sectoral and geographical contributions to summertime US black carbon (BC) distributions are studied. NA emissions heavily (>70%) affect the BC levels from the surface to 5 km, while non-NA plumes compose more than half of the BC above 5 km. NA and non-NA BB, NA transportation and non-NA residential emissions are the major contributing sectors. Aircraft measurements during the California phase of the Arctic Research of the Composition of the Troposphere from Aircraft andSatellites (ARCTAS-CARB) field campaign show that BC/(organic matter + nitrate + sulfate) mass ratios fairly well represent BC's warming potential over southern California, which can be approximated by BC/(organic matter + sulfate) and BC/sulfate for plumes affected and unaffected by fires, respectively. The responses of BC/(organic matter + sulfate) and BC/sulfate to removing each emission sector indicate that mitigating NA transportation emissions has the highest potential for regional air quality and climate co-benefits. Contributions from NA BB and extra-regional emissions differ for summer and spring (April 2008).

The proliferation of multimedia data, including digital images and videos, has led to an increase in their misuse, such as the unauthorized sharing of sensitive content, the spread of fake news, and the dissemination of misleading propaganda. To address these issues, the research field of multimedia forensics has developed tools to distinguish genuine multimedia from fakes and identify the sources of those who share sensitive content. However, the accuracy and reliability of multimedia forensics tools are threatened by recent technological advancements in new multimedia processing software and camera devices. For example, source attribution involves attributing an image or video to a specific camera device, which is crucial for addressing privacy violations, cases of revenge porn, and instances of child pornography. These tools exploit forensic traces unique to each camera’s manufacturing process, such as Photo Response Non-Uniformity (PRNU). Nevertheless, image and video processing transformations can disrupt the consistency of PRNU, necessitating the development of new methods for its recovery. Conversely, to distinguish genuine multimedia from fakes, AI-based image and video forgery localization methods have also emerged. However, they constantly face challenges from new, more sophisticated AI-forgery techniques and are hindered by factors like AI-aided post-processing and, in the case of videos, lower resolutions, and stronger compression. This doctoral study investigates the relevance of exploiting temporal information during the parameters estimation used to reverse complex spatial transformations for source attribution, and video forgery localization in low-resolution H.264 post-processed inpainted videos. Two novel methods will be presented that model the set of parameters involved in reversing in-camera and out-camera complex spatial transformations applied to images and videos as time series, improving source attribution accuracy and computational efficiency. Regarding video inpainting localization, a novel dataset of videos inpainted and post-processed with Temporal Consistency Networks will be introduced, and we will present our solution to improve video inpainting localization by taking into account spatial and temporal inconsistencies at dense optical flow level. The research presented in this dissertation has resulted in several publications that contribute to the field of multimedia forensics, addressing challenges related to source attribution and video forgery localization.

This research investigates the application of machine learning techniques in computational genomics across two distinct domains: (1) the predicting the source of bacterial pathogen using whole genome sequencing data, and (2) the functional annotation of genes using single- cell RNA sequencing data. This work proposes the development of a bioinformatics pipeline tailored for identifying genomic variants, including gene presence/absence and single nu- cleotide polymorphism. This methodology is applied to specific strains such as Salmonella enterica serovar Typhimurium and the Ralstonia solanacearum species complex. Phylo- genetic analyses along with pan-genome and positive selection studiesshow that genomic variants and evolutionary patterns of S. Typhimurium vary across sources, which suggests that sources can be accurately attributed based on genomic variants empowered by machine learning. We benchmarked seven traditional machine learning algorithms, achieving a no- table accuracy of 94.6% in host prediction for S. Typhimurium using the Random Forest model, underscored by SHAP value analyses which elucidated key predictive features. Next, the focus is shifted to the prediction of Gene Ontology terms for Arabidopsis genes using single-cell RNA-seq data. This analysis offers a detailed comparison of gene expression in root versus shoot tissues, juxtaposed with insights from bulk RNA-seq data. The integration of regulatory network data from DAP-seq significantly enhances the prediction accuracy of gene functions. / Master of Science / This work applies machine learning techniques to two areas in computational biology: pre- dicting the hosts of bacterial pathogens based on their genome data, and predicting the func- tions of plant genes using single-cell gene expression data. The first part develops a method to analyze genome sequences from bacterial pathogens like Salmonella enterica serovar Ty- phimurium and the Ralstonia solanacearum species complex, identifying genomic variants, including gene presence/absence and single nucleotide polymorphism, which are variations in genetic code. By studying the evolutionary relationships and genetic diversity among dif- ferent strains, the motivation for using machine learning models to predict the sources (e.g., poultry, swine) of the pathogen genomes is established. Several machine learning models are then trained on these datasets, and the most important factors contributing to the predic- tions are identified. The second part focuses on predicting the functions of genes in the model plant species Arabidopsis thaliana using the gene expression data measured at the single-cell level to train machine learning models for identifying standardized gene function descrip- tions called Gene Ontology (GO) terms. By comparing results from single-cell and bulk tissue data, the study evaluates whether the higher resolution of single-cell data improves gene function prediction accuracy. Additionally, by incorporating information about gene regulation from a specialized experiment, the role of gene expression control in determining gene functions is explored.

Rapid and reliable techniques are necessary for the analysis of accelerants, including gasoline, from fire debris evidence in forensic arson investigations. Gasoline additives can be used as chemical attribute signatures (CAS) to distinguish between source locations due to the variation in additives used. Source attribution using CAS is needed in forensic chemistry, as the determination of a single gasoline source could be a potential investigation tool for law enforcement and other agencies conducting arson investigation. Direct analysis in real time-mass spectrometry (DART-MS) has had increasing popularity in the field of forensic chemistry for chemical analysis, and it has been applied to fire debris analysis. DART-MS has great capacity for gasoline source attribution due to its ionization technique and inclusion of higher molecular weight ions, which correspond to the CAS in gasoline. To test the hypothesis of gasoline source attribution, 21 gasoline samples were collected across Massachusetts, New Hampshire, and Connecticut. DART-MS data were generated for each sample of gasoline in replicates of 10. The data were grouped based on geographical location and evaluated by Principal Component Analysis (PCA). PCA was used to evaluate the similarities and differences in gasoline DART-MS data by generating and classifying the gasoline sample groups formed. Leave-one-out cross-validation (LOOCV) was performed on each geographical group after PCA. LOOCV was used as the validation technique to determine the validity of the model and asses its capability at classifying unknown gasoline samples. DART-MS data across geographical groups was found to have varying levels of similarity and difference through visual inspection of the mass spectra. PCA showed distinct groupings of individual gasoline samples across all tested geographical groups, with 3 out of 6 geographical groups showing no overlap between gasoline sample classifications. Two groups showed minimal overlapping, while 1 group had overlapping between multiple gasoline sample classifications. Three groups had a LOOCV of 100% with no misclassifications. The other LOOCV were 98%, 96.67%, and 85%. The PCA and comparison of DART-MS data provides evidence of successful differentiation between gasoline samples of the same brand across Massachusetts, New Hampshire, and Connecticut. This research aims to provide an overview and understanding of chemometrics and DART-MS and how these techniques may be applied for forensic source attribution purposes.

Campylobacter est responsable de la zoonose bactérienne d’origine alimentaire la plus fréquemment reportée en Europe. Cette bactérie étant ubiquitaire, les sources et voies d’infection de l’Homme sont nombreuses. Cependant, afin de diminuer l’incidence de la maladie, il est nécessaire d’identifier les principaux réservoirs impliqués dans les infections humaines. Pour cela, nous avons dans un premier temps investigué la présence de Campylobacter dans trois réservoirs animaux (volaille, bovin, animaux de compagnie), ainsi que la diversité génétique des isolats de C. jejuni, en comparaison à celle d’isolats cliniques, à l’aide des techniques MLST (Multilocus sequence typing) et CGF (Comparative Genomic Fingerprinting). Afin d’identifier l’origine des campylobactérioses avec précision et de compenser notamment les limites techniques de la MLST, 15 marqueurs génétiques ont été sélectionnés comme marqueurs potentiellement indicateurs de l’hôte, après analyse de plus de 800 génomes de C. jejuni. Par la suite, la capacité de la MLST, la CGF40 et des 15 marqueurs à identifier l’origine des campylobactérioses a été étudiée. Ainsi, les 15 marqueurs se sont révélés être particulièrement performants pour l’attribution de sources des campylobactérioses, suivis ensuite par la MLST, tandis que la CGF40 est apparue comme étant peu adaptée. A partir des données MLST et des 15 marqueurs génétiques, une implication majoritaire des volailles et des bovins a été mis en évidence en France, tandis que les animaux de compagnie et l’environnement (comprenant eau et oiseaux sauvages) étaient faiblement impliqués. Ceci permet ainsi de renforcer les efforts de recherche relatifs aux moyens de lutte contre Campylobacter menés dans ces réservoirs. Ce travail a également permis de mettre en évidence de potentielles spécificités nationales dans la dynamique de transmission de C. jejuni à l’Homme. / Campylobacter is the causal agent of the main bacterial foodborne gastroenteritis in Europe. Since Campylobacter is frequently found in animal reservoirs, sources of human infection and transmission routes are various. However, to decrease the human burden of campylobacteriosis, it is essential to quantify the relative importance of the several reservoirs in human infections. For this purpose, we assessed the contamination of chicken, cattle and pets by Campylobacter spp., and further characterized C. jejuni isolates using MLST (Multilocus Sequence Typing) and CGF (Comparative Genomic Fingerprinting) in comparison with French clinical isolates. Then, in order to identify the most likely origin of campylobacteriosis cases in France and overcome MLST limitations in source attribution, about 800 C. jejuni genomes were analyzed which resulted in the identification of 15 genes as promising host segregating markers for source attribution. Subsequently, we assessed the ability of MLST, CGF40 and the 15 host-segregating markers to identify the most likely origin of campylobacteriosis. The 15 host-segregating markers were the most powerful in source attribution, followed by MLST, while CGF40 appeared to be not suitable for source attribution in our study. Based on MLST and the 15 markers, assignments of clinical cases emphasize the significant implication of chicken and ruminant in human infection by Campylobacter, while pets and the environment (including water and wild birds) were slightly involved, reinforcing the interest to focus control strategies on livestock. Finally this work highlights potential national variations in the transmission dynamics of C. jejuni to human.

When intelligent voice-based assistants (VBAs) present news, they simultaneously act as interlocutors and intermediaries, enabling direct and mediated communication. Hence, this study discusses and investigates empirically how interlocutor and intermediary predictors affect an assessment that is relevant for both: trustworthiness. We conducted a secondary analysis using data from two online surveys in which participants (N = 1288) had seven quasi-interactions with either Alexa or Google Assistant and calculated hierarchical regression analyses. Results show that (1) interlocutor and intermediary predictors influence people’s trustworthiness assessments when VBAs act as news presenters, and (2) that different trustworthiness dimensions are affected differently: The intermediary predictors (information credibility; company reputation) were more important for the cognition-based trustworthiness dimensions integrity and competence. In contrast, intermediary and interlocutor predictors (ontological classification; source attribution) were almost equally important for the affect-based trustworthiness dimension benevolence.