<p dir="ltr">Antigenically variable pathogens (AVPs) pose a significant infectious disease burden, but vaccine development is extremely difficult due to their ability to quickly evolve beyond host immunity. Existing models of AVP spread have not been able to sufficiently account for host immune history, population mobility patterns, and pathogen evolutionary dynamics. This thesis aims at creating a computational framework built from the concept of antibody landscapes to overcome these issues, thereby increasing the understanding of how these pathogens spread and evolve in order to improve vaccine design.</p><p><br></p><p dir="ltr">Briefly, the proposed stochastic framework is built from "the ground up'' using principles of antibody landscapes, in which we begin by devising a mechanism to describe how the landscape changes due to repeated pathogen exposure. Extending this to a (sub)population-level permits integration into a meta-population model that is further parameterized by geographic influences. Virus evolution is driven by a statistically meaningful model of antigenic drift in the underlying antigenic space. While the framework is robust and, in principle, capable of modeling a variety of AVPs, we focus on influenza H3N2 as a case study due to its data availability and persistently low and unpredictable vaccine efficacy.</p><p><br></p><p dir="ltr">Experimental results demonstrate that we can statistically significantly predict various properties of H3N2 evolution and population level immunity, including prevalence level, the timing of emergence of new antigenic clusters, the positions of unseen strains in antigenic space, as well as the geographic locations where new strains and antigenic clusters emerge. Through analysis of the simulated outcomes, we identified a population level of immune protection against circulating strains (titre value of approximately 5 units), which when approached, seems to signal an upcoming antigenic drift. Using this insight, we propose a new vaccine strain selection strategy that shows notable improvements in vaccine effectiveness and stability. Additionally, we estimate that it could reduce annual morbidity by 73.4 ± 40.8 million (17% ± 9%) in the Northern Hemisphere and 56.7 ± 38.0 million (10% ± 6%) in the Southern Hemisphere. In summary, this novel framework can accurately replicate the interplay between pathogen evolution and population-level immune responses decades into the future from a mechanistic perspective, and be used to design improved vaccines.</p>

La presente tesi intende esplorare gli effetti della resistenza parziale sulle epidemie di peronospora della vite (Plasmopara viticola). È stato sviluppato un modello di simulazione teorico che comprende lo sviluppo della pianta ospite e le fasi principali della malattia, dalla mobilizzazione dell’inoculo, alla moltiplicazione della malattia sulle foglie, all’infezione dei grappoli. Attraverso esperimenti (monociclici) di inoculazione è stata studiata la risposta alle infezioni di P. Viticola di 16 varietà parzialmente resistenti, analizzando le seguenti componenti: frequenza d’infezione, durata del periodo di latenza, dimensione delle lesioni, produzione di sporangi, durata del periodo infezioso e infettività degli sporangi prodotti sulle lesioni. Queste componenti di resistenza sono state incorporate nel modello, attraverso cui sono stati studiati i loro effetti sull’epidemia (policiclica) in diversi scenari. Le componenti di resistenza hanno mostrato diversi livelli di efficacia nel sopprimere l’epidemia: l’efficienza di infezione e la produzione di sporangi risultano avere un maggiore impatto nella resistenza espressa a livello di pieno campo. Questo approccio è utile per guidare lo studio fenotipico della resistenza dell’ospite e per anticipare le prestazioni di un genotipo a livello di pieno campo, che risulterebbe difficile e dispendioso considerando la natura perenne della vite. / The present dissertation aims to explore the effects of partial resistance on grapevine downy mildew (Plasmopara viticola) epidemics. A theoretical simulation model was developed including host dynamics and main phases of the disease, from inoculum mobilisation to disease multiplication on foliage, and to infection of clusters. The response to P. Viticola infection was studied for 16 grapevine varieties through (monocyclic) inoculation experiments, by measuring components of partial resistance: infection frequency, duration of latent period, size of lesions, production of sporangia, duration of infectious period, and infectivity of sporangia produced on lesion. Components of partial resistance were incorporated into the model and their effects on the (polycyclic) epidemic were investigated accross different scenarios. Components of partial resistance showed different effectiveness on the suppression of epidemics, infection efficiency and spore production having the strongest impact on the overall field resistance response. This approach is an useful tool for phenotyping studies on host plant resistance and for anticipating the performance of a genotype at the field scale, that otherwise is difficult and time requiring due to the perennial nature of grapevine.

La paratuberculosis (PTB) est une maladie endémique des ruminants causée par Mycobacterium avium subsp. paratuberculosis (Map). Les stratégies de maîtrise actuelles ne sont pas suffisamment efficaces. La réponse à l'exposition à Map varie entre les animaux avec une part de déterminisme génétique. Des marqueurs génétiques pourraient permettre une sélection. L'objectif était d'évaluer par modélisation l'efficacité potentielle attendue de stratégies de maîtrise utilisant la sélection génétique ou la réduction de l'exposition en élevage. Nous avons identifié quatre traits phénotypiques de résistance influençant principalement la propagation de Map à l'échelle du troupeau et montré la valeur ajoutée de leur amélioration simultanée. Nous avons évalué l'effet de l'environnement du troupeau et du système d’élevage sur la propagation et la maîtrise de Map. Nous avons montré une différence d’efficacité des stratégies de maîtrise les plus pertinentes entre deux systèmes d'élevage bovins laitiers contrastés d'Europe: l'ouest de la France et l'Irlande. Nous avons évalué l'efficacité que pourrait apporter la sélection génomique en évaluant le temps nécessaire pour atteindre des niveaux de variation des traits sélectionnés permettant un bon contrôle de l‘infection sous l’hypothèse que des marqueurs de sélection soient disponibles. Nous avons identifié 2 paramètres du modèle de sélection génomique influents sur l’efficacité de la sélection. Notre modèle permet d’intégrer de nouvelles connaissances biologiques sur le déterminisme génétique de la résistance à Map pour évaluer des stratégies de maîtrise complexes comprenant une composante de sélection génomique. / Paratuberculosis (PTB) is an endemic disease of ruminants caused by Mycobacterium avium subsp. paratuberculosis (Map). Current control strategies are not effective enough. The response to Map exposure varies between animals with evidence of a partial genetic determinism. Genetic markers could allow selection. The objective was to assess the potential expected effectiveness of control strategies relying on genetic selection or reduction of exposure in herds, using a modelling approach. We identified four phenotypic traits of resistance mainly influencing the spread of Map at the herd scale and showed the added value of their simultaneous improvement. We evaluated the effect of the herd environment and management on the spread and control of Map. We showed a difference in effectiveness of the most relevant control strategies between two contrasting dairy cattle systems in Europe: western France and Ireland. We evaluated the effectiveness of genomic selection by assessing the time required to reach levels of variation in the selected traits allowing to achieve a good control of infection, assuming that associated genomic markers could be available. Effectiveness of selection was mainly influenced by 2 of the parameters of the developed genomic selection model. Our model allows to account for future knowledge about the genetic determinism of cattle resistance to Map in order to assess the effectiveness of complex control strategies including a genomic selection component.

La modélisation mathématique est largement utilisée pour effectuer des recherches sur la modélisation des maladies infectieuses. Combler le fossé entre les modèles conceptuels et leurs simulations est l'un des problèmes de la modélisation. Les langages métiers sont souvent utilisés pour addresser ces problèmes en séparant deux aspects de la modélisation : la spécification (modèles conceptuels) et la simulation (modèles informatiques). Dans cette perspective, nous développons un langage métier, appelé KENDRICK, dédié à la modélisation épidémiologique, couplé avec une plate-forme de simulation. Un autre problème de la modélisation en épidémiologie est le mélange des aspects de domaine qui doivent être séparés. Afin de faciliter l'écriture et l'évolution des modèles, il est crucial de pouvoir définir une préoccupation avec aussi peu de dépendances avec d'autres que possible et de pouvoir les combiner aussi librement que possible. Nous abordons ces défis en proposant un méta-modèle mathématique commun qui peut représenter les modèles ainsi que les préoccupations. Nous définissons ensuite les opérateurs qui permettent de combiner des préoccupations ainsi que de les appliquer dans un modèle. Le langage KENDRICK simplifie donc la programmation des simulations épidémiologiques en décomposant un modèle monolithique hautement-couplé en préoccupations modulaires. Cela rend alors plus facile la construction des modèles complexes de l'épidémiologie où plusieurs préoccupations sont considérées en même temps. / Mathematical and computational models have become widely used and demanded tools for examining mechanisms of transmission, exploring characteristics of epidemics, predicting future courses of an outbreak and evaluating strategies to find a best control-program. One of the problems of modelling is bridging the gap between conceptual models (i.e compartmental models of epidemiology) and their computer simulation (through deterministic, stochastic or agent-based implementation). Domain Specific Languages (DSLs) are often used to address such difficulties by separating two concerns of modelling, specification (conceptual model) and implementation (computational model). In this perspective, we develop a DSL called KENDRICK targeted to the epidemiological modelling and coupled with a simulation platform that allows the study of such models. The other important issue needs to be addressed in the context of epidemiological modelling is the heterogeneities introduced by separate concerns. In order to facilitate the specification of models and their evolution, it is crucial to be able to define concerns with as few dependencies with each other as possible and to combine them as freely as possible. We address such challenges by proposing a common mathematical meta-model that supports both concerns and models and enabling their compositions by some operators. We then implement our proposal language KENDRICK based on this meta-model. The language simplifies the construction of complex epidemiological models by decomposing them into modular concerns, by which common concerns can be reused across models and can be easily changed.

A modelagem epidemiológica é uma importante ferramenta que auxilia os órgãos de saúde no controle de doenças infecciosas, pois permitem analisar e comparar diversas estratégias que facilitam a tomada de decisões e definições de protocolos. A dengue é atualmente a doença viral humana com maior número de casos. Possui índice de mortalidade baixo, entretanto, é endêmica em mais de 100 países e 40% da população mundial está em risco de contrair a infecção. Através dos dados de notificação de dengue no Brasil, evidenciamos que os surtos são sazonais, que há alternância de sorotipos ao longo dos anos e mostramos que a doença é diferente em cada localização, e que somente com uma normalização adequada é possível sugerir um agrupamento coerente de municípios. Neste trabalho, as informações obtidas a partir dos dados são usadas para a estruturação dos modelos matemáticos e para a estimação de parâmetros que validam estes modelos. Comparamos a dinâmica de transmissão de dengue do modelo com um sorotipo, com modelos que permitem a interação de dois, três e quatro sorotipos simultaneamente, além da possibilidade de até quatro infecções sequenciais. Os modelos com múltiplos sorotipos são expandidos do modelo básico que categoriza hospedeiros dentro de uma população como suscetíveis (S), infectados (I) e recuperados (R) e acoplado à dinâmica dos vetores suscetíveis (V) e infectados (Vi). Nossos modelos incluem: um período de imunidade cruzada de forma que o indivíduo adquire imunidade permanente para o sorotipo que já foi infectado e imunidade temporária para os demais; uma forçante de sazonalidade na taxa de nascimento dos vetores; uma assimetria com taxas de transmissão diferentes para cada sorotipo; e o compartimento dos vacinados, com uma vacina tetravalente que confere diferentes imunidades para cada sorotipo. Os resultados mostram que para a reprodução de surtos anuais é necessário a inclusão da forçante de sazonalidade na taxa de nascimento dos vetores, e que o modelo com quatro sorotipos é o que melhor reproduz os dados de incidência de dengue, sendo o mais adequado para analisar estratégias de vacinação com uma vacina tetravalente. Comparamos duas estratégias de vacinação: vacinação aleatória na população e vacinação direcionada para faixas etárias. Neste caso, os resultados demonstram a superioridade da estratégia direcionada e que as escolhas das faixas etárias devem ser definidas por município e não por um protocolo nacional. / Epidemiological modelling is an important tool that assists the health agencies in the control of infectious diseases, since it allows analysing and to compare several strategies that facilitate decision-making and protocol definitions. Dengue is currently the most important vector-borne disease. The mortality rate of dengue is low, however, it is endemic in more than 100 countries and about 40% of the world\'s population is at risk of contracting the infection. Through the dengue notification data in Brazil, we emphasize that the outbreaks are seasonal, there is serotypes alternation over the years and we show that the disease is different in each locality, and that only with a suitable standardization it is possible to propose an appropriate grouping of municipalities. In this work, we use the data information to formulate the mathematical models and for the parameter\'s estimation in order to validate these models. We compare the dynamics of dengue of the one serotype model with the models that allow interaction of two, three and four serotypes simultaneously, including the possibility of at most four sequential infections.The multi-strain models are expanded from the basic model which categorizes the host population as susceptible (S), infected (I), and recovered (R) and coupled with the dynamics of the susceptible (V) and infected (Vi) vectors. Our models include: a period of cross-immunity which means permanent immunity to the serotype of the infection and temporary immunity to the other serotypes; a seasonal forcing in the mosquitoes birth rate; different transmissions rates, so that the models are asymmetric; and the compartment of vaccinated individuals with a tetravalent vaccine which confers different immunities for each serotype. The results show that to reproduce yearly outbreaks it is necessary to include the seasonal forcing in the birth rate of the vectors, and that the four serotypes model is the one that best reproduces the dengue incidence data, being the most suitable model to analyse vaccination strategies with a tetravalent vaccine. We compare two vaccination strategies: random vaccination and vaccination targeted at age groups. In this case, the results demonstrate the superiority of the targeted strategy and that the choices of the age groups should be defined by municipality and not by a national protocol.

The last two decades have seen several large-scale epidemics of international impact, including human, animal and plant epidemics. Policy makers face health challenges that require epidemic predictions based on limited information. There is therefore a pressing need to construct models that allow us to frame all available information to predict an emerging outbreak and to control it in a timely manner. The aim of this thesis is to develop an early-warning modelling approach that can predict emerging disease outbreaks. Based on Bayesian techniques ideally suited to combine information from different sources into a single modelling and estimation framework, I developed a suite of approaches to epidemiological data that can deal with data from different sources and of varying quality. The SEIR model, particle filter algorithm and a number of influenza-related datasets were utilised to examine various models and methodologies to predict influenza outbreaks. The data included a combination of consultations and diagnosed influenza-like illness (ILI) cases for five influenza seasons. I showed that for the pandemic season, different proxies lead to similar behaviour of the effective reproduction number. For influenza datasets, there exists a strong relationship between consultations and diagnosed datasets, especially when considering time-dependent models. Individual parameters for different influenza seasons provided similar values, thereby offering an opportunity to utilise such information in future outbreaks. Moreover, my findings showed that when the temperature drops below 14°C, this triggers the first substantial rise in the number of ILI cases, highlighting that temperature data is an important signal to trigger the start of the influenza epidemic. Further probing was carried out among Maltese citizens and estimates on the under-reporting rate of the seasonal influenza were established. Based on these findings, a new epidemiological model and framework were developed, providing accurate real-time forecasts with a clear early warning signal to the influenza outbreak. This research utilised a combination of novel data sources to predict influenza outbreaks. Such information is beneficial for health authorities to plan health strategies and control epidemics.

A modelagem epidemiológica é uma importante ferramenta que auxilia os órgãos de saúde no controle de doenças infecciosas, pois permitem analisar e comparar diversas estratégias que facilitam a tomada de decisões e definições de protocolos. A dengue é atualmente a doença viral humana com maior número de casos. Possui índice de mortalidade baixo, entretanto, é endêmica em mais de 100 países e 40% da população mundial está em risco de contrair a infecção. Através dos dados de notificação de dengue no Brasil, evidenciamos que os surtos são sazonais, que há alternância de sorotipos ao longo dos anos e mostramos que a doença é diferente em cada localização, e que somente com uma normalização adequada é possível sugerir um agrupamento coerente de municípios. Neste trabalho, as informações obtidas a partir dos dados são usadas para a estruturação dos modelos matemáticos e para a estimação de parâmetros que validam estes modelos. Comparamos a dinâmica de transmissão de dengue do modelo com um sorotipo, com modelos que permitem a interação de dois, três e quatro sorotipos simultaneamente, além da possibilidade de até quatro infecções sequenciais. Os modelos com múltiplos sorotipos são expandidos do modelo básico que categoriza hospedeiros dentro de uma população como suscetíveis (S), infectados (I) e recuperados (R) e acoplado à dinâmica dos vetores suscetíveis (V) e infectados (Vi). Nossos modelos incluem: um período de imunidade cruzada de forma que o indivíduo adquire imunidade permanente para o sorotipo que já foi infectado e imunidade temporária para os demais; uma forçante de sazonalidade na taxa de nascimento dos vetores; uma assimetria com taxas de transmissão diferentes para cada sorotipo; e o compartimento dos vacinados, com uma vacina tetravalente que confere diferentes imunidades para cada sorotipo. Os resultados mostram que para a reprodução de surtos anuais é necessário a inclusão da forçante de sazonalidade na taxa de nascimento dos vetores, e que o modelo com quatro sorotipos é o que melhor reproduz os dados de incidência de dengue, sendo o mais adequado para analisar estratégias de vacinação com uma vacina tetravalente. Comparamos duas estratégias de vacinação: vacinação aleatória na população e vacinação direcionada para faixas etárias. Neste caso, os resultados demonstram a superioridade da estratégia direcionada e que as escolhas das faixas etárias devem ser definidas por município e não por um protocolo nacional. / Epidemiological modelling is an important tool that assists the health agencies in the control of infectious diseases, since it allows analysing and to compare several strategies that facilitate decision-making and protocol definitions. Dengue is currently the most important vector-borne disease. The mortality rate of dengue is low, however, it is endemic in more than 100 countries and about 40% of the world\'s population is at risk of contracting the infection. Through the dengue notification data in Brazil, we emphasize that the outbreaks are seasonal, there is serotypes alternation over the years and we show that the disease is different in each locality, and that only with a suitable standardization it is possible to propose an appropriate grouping of municipalities. In this work, we use the data information to formulate the mathematical models and for the parameter\'s estimation in order to validate these models. We compare the dynamics of dengue of the one serotype model with the models that allow interaction of two, three and four serotypes simultaneously, including the possibility of at most four sequential infections.The multi-strain models are expanded from the basic model which categorizes the host population as susceptible (S), infected (I), and recovered (R) and coupled with the dynamics of the susceptible (V) and infected (Vi) vectors. Our models include: a period of cross-immunity which means permanent immunity to the serotype of the infection and temporary immunity to the other serotypes; a seasonal forcing in the mosquitoes birth rate; different transmissions rates, so that the models are asymmetric; and the compartment of vaccinated individuals with a tetravalent vaccine which confers different immunities for each serotype. The results show that to reproduce yearly outbreaks it is necessary to include the seasonal forcing in the birth rate of the vectors, and that the four serotypes model is the one that best reproduces the dengue incidence data, being the most suitable model to analyse vaccination strategies with a tetravalent vaccine. We compare two vaccination strategies: random vaccination and vaccination targeted at age groups. In this case, the results demonstrate the superiority of the targeted strategy and that the choices of the age groups should be defined by municipality and not by a national protocol.