We present a variant of the Q-learning algorithm with automatic control of the exploration rate by a competition scheme. The theoretical approach is accompanied by systematic simulations of a chaos control task. Finally, we give interpretations of the algorithm in the context of computational ecology and neural networks.

Bacteriophages (viruses that infect bacteria) are the most abundant biological life-forms on Earth. However, very little is known regarding the structure of phage-bacteria infections. In a recent study we showed that phage-bacteria infection assay datasets are statistically nested in small scale communities while modularity is not statistically present. We predicted that at large macroevolutionary scales, phage-bacteria infection assay datasets should be typified by a modular structure, even if there is nested structure at smaller scales. We evaluate and confirm this hypothesis using the largest study of the kind to date. The study in question represents a phage-bacteria infection assay dataset in the Atlantic Ocean region between the European continental shelf and the Sargasso Sea. We present here a digitized version of this study that consist of a bipartite network with 286 bacteria and 215 phages including 1332 positive interactions, together with an exhaustive structural analysis of this network. We evaluated the modularity and nestedness of the network and its communities using a variety of algorithms including BRIM (Bipartite, Recursively Induced Modules), NTC (Nestedness Temperature Calculator) and NODF (Nestedness Metric based on Overlap and Decreasing Filling). We also developed extensions of these standard methods to identify multi-scale structure in large phage-bacteria interaction datasets. In addition, we performed an analysis of the degree of geographical diversity and specialization among all the hosts and phages. We find that the largest-scale ocean dataset study, as anticipated by Flores et al. 2013, is highly modular and not significantly nested (computed in comparison to null models). More importantly is the fact that some of the communities extracted from Moebus and Nattkemper dataset were found to be nested. We examine the role of geography in driving these modular patterns and find evidence that phage-bacteria interactions can exhibit strong similarity despite large distances between sites. We discuss how models can help determine how coevolutionary dynamics between strains, within a site and across sites, drives the emergence of nested, modular and other complex phage-bacteria interaction networks. Finally, we releases a computational library (BiMAT)to help to help the ecology research community to perform bipartite network analysis of the same nature I did during my PhD.

<p dir="ltr">Metabolic pathways are shaped by dynamic biotic interactions. My research delves into coevolution exemplified through two distinct projects that investigate the specialized metabolism of organisms as a consequence of biotic interactions. The first project focused on the remarkable metabolic adaptations of <i>Elysia crispata</i> morphotype clarki. This sea slug possesses the extraordinary ability to sequester and maintain functional chloroplasts (kleptoplasts) from the algae it consumes, allowing it to sustain photosynthetically active kleptoplasts for several months without feeding. To better understand the underlying molecular mechanism of this phenomenon, I generated a comprehensive 786 Mbp draft genome of <i>E. crispata</i> using a combination of ONT long reads and Illumina short reads. The resulting assembly provided a foundational resource for phylogenetic, gene family and gene expression analyses. This work advanced our understanding of the genetic underpinnings of kleptoplasty, shedding light on the evolution and maintenance of this unique metabolic strategy in sacoglossan sea slugs. I next investigated the transcriptional impacts of herbivory on maize (<i>Zea mays</i>) and green foxtail (<i>Setaria viridis</i>), induced by fall armyworm (<i>Spodoptera frugiperda</i>) and beet armyworm (<i>Spodoptera exigua</i>) feeding. This study aimed to contrast the defensive mechanisms of these grasses in response to each herbivore, and determined that green foxtail transcriptionally differentiates its responses to fall armyworm and beet armyworm herbivory. The fall armyworm has evolved a counter adaptation to lessen plant secondary metabolite production by producing a salivary protein (SFRP1) that suppresses jasmonate signaling. Investigation of the combinatorial effects of SFRP1 and beet armyworm herbivory determined the addition of endogenous SFRP1 during beet armyworm feeding is sufficient to reduce green foxtail defense responses. Results of this research shed light on host-pest reciprocal adaptations and the role of SFRP1 as an oral secretory protein. Coexpression analysis of maize and green foxtail transcriptomic responses to herbivory also identified putative genes involved in specialized metabolic pathways in green foxtail, providing insights into plant-insect interactions and potential solutions to herbivory in wild plant species. These findings highlight how gene diversification can contribute to pest resistance in grasses. Together, these seemingly unconnected projects underscore how biotic interactions influence metabolic processes across diverse organisms and reveal the fascinating intricacies of their adaptations to environmental challenges.</p>

Phytoplankton are single celled organisms capable of phytosynthesis, and are present in all the major oceans and lakes in the world. Phytoplankton contribute to 50% of the total primary production on Earth, and are the dominating primary producer in most aquatic ecosystems. This thesis is based on the 1D deterministic model by Jäger et. al. (2010) which models phytoplankton dynamics in freshwater lakes, where phytoplankton growth is limited by the availability of light and phosphorus. The original model is here extended to two dimensions to include a horizontal dimension as well as a vertical dimension, in order to simulate phytoplankton dynamics under varying lake bottom topographies. The model was solved numerically using a grid transform and a finite volume method in MATLAB. Using the same parameter settings as the 1D case studied by Jäger et. al. (2010), an initial study of plankton dynamics was done by varying the horizontal and vertical diffusion coefficients independently.

The coexistence of species is probably one of the most interesting and complex phenomenon in nature. We constructed an agent based model to study the coexistence dynamics of prey - predator populations by varying productivity levels of producers in fragmented and connected habitats along with different levels of quality of predators. Our results indicated that productivity levels of producers in fragmented and connected habitats along with levels of predator quality are significantly responsible for overall predator - prey population size and survivorship. In the absence of predation, competition between identical prey populations is more probable in connected habitats than in unfragmented or fragmented habitats. Implementing low quality predators in the habitats positively influences the overall coexistence dynamics whereas implementing high quality predators tend to decrease the prey populations. Fragmented habitats provide for greater prey population survival time in highly productive environments but low prey population survival time in less productive environments.

<p dir="ltr">In a rapidly changing world, human actions and natural events are reshaping ecosystems and presenting new challenges for conservation efforts. Within this context, unraveling the recent ecosystem transformations and their implications on a fine scale is required. The impacts of such changes are not always sudden but often gradual and sometimes as a result of historical events. With the recent advancement in technologies, the resolution of information by genome sequences spans from millions of years ago (hindcasting) to future generations (forecasting). Aquatic ecosystems pose their own challenges when it comes to ecosystem changes and the types of data required to assess impact and help inform conservation efforts. My dissertation comprises three chapters focused on using genomic techniques to generate data valuable for the conservation and management of aquatic ecosystems. Each of the three chapters is a distinct manuscript in terms of scientific publications, where Chapter 1 has already been published, Chapter 2 has been submitted to a journal, revised, and is now awaiting publication, and Chapter 3 is in preparation for submission to a peer-reviewed journal. In Chapter 1, dietary DNA from harvested North American river otter (<i>Lontra canadensis</i>)<i> </i>was used to determine whether metabarcoding of stomach content could be used to identify fish prey species consumed. In Chapter 2, DNA sequencing of endangered pupfish species in the Tularosa Basin of New Mexico was studied; before my work, it was nominally comprised of a single species, the White Sands pupfish (<i>Cyprinodon tularosa</i>). The results indicate a rapid speciation event occurred within about the last ~5000 years, driven primarily by genetic drift. Chapter 3 extends Chapter 2 by assessing the dynamics of genomic diversity over space and time while evaluating the short-term evolutionary dynamics (~18 generations) of the two native pupfish populations. This temporal study aimed to determine if the extraordinarily rapid evolution over the last ~5000 years (observed in Chapter 2) could be detected over timescales more relevant to conservation and management efforts. Overall, this dissertation used genomic sequence data from metabarcoding of the COI gene region in the otter stomach content as well as pool sequencing and whole genome resequencing of pupfish to provide key biological insights into the conservation of these aquatic species. This dissertation also provides insights into avenues for further study and highlights the significant role that conservation genomics can play in the future. The findings presented in the three chapters are discussed within the context of species’ conservation and management.</p>

Les interactions font partie intégrante des écosystèmes. Que ce soit aux niveaux les plus fins, comme les protéines, ou les plus larges, comme les méta-communautés, il est possible de les regrouper en réseaux et d’en étudier la structure. Cela a permis de mettre en évidence que certaines structures sont observables à différents niveaux, c’est le cas par exemple des réseaux emboîtés. De plus, les réseaux d’interaction ont la spécificité de ne pas être fixes dans le temps et l’espace, ce qui leur confère un avantage de taille pour l’étude de l’évolution. Ils peuvent ainsi servir de support à l’études des mécanismes intervenants dans les processus évolutifs. Cependant, il n’existe pas encore de méthodologie ayant fait consensus sur l’utilisation des réseaux et leur analyse à différentes échelles d’organisation. Cette thèse se base sur l’hypothèse que les réseaux, de par leurs propriétés, sont pertinents à considérer pour comprendre l’évolution et ce à différentes échelles d’organisation, et offrent la possibilité de faire des liens entre chacune d’entre elles. L’approche basée sur les réseaux, combinée à l’utilisation de modèles théorique serait donc un outil méthodologique puissant dans l’élargissement des connaissances concernant les processus sous-jacents à l’évolution. La thèse qui suit composée de six chapitres dont le contenu est le suivant. Elle commence par un chapitre d’introduction aux concepts d’intérêts, notamment sur l’évolution et la coévolution. Le deuxième chapitre est une introduction à l’utilisation des réseaux en écologie, suivit par le troisième chapitre qui effectue une revue non exhaustive des méthodologies développées autour des réseaux d’interactions. Les chapitres suivants sont en quelque sorte une mise en pratique de ces méthodes et ce à différents niveaux d’organisation. Le quatrième chapitre revient sur une étape avortée de ce doctorat qui servira tout de même à la construction du modèle du chapitre suivant. Le cinquième chapitre se concentre sur la coévolution et son suivit au travers des réseaux d’interaction entre les bactéries et leurs virus. Enfin, le sixième chapitre traque l’évolution des communautés grâce à la structure des arbres phylogénétiques et structure des réseaux d’interactions au cours du temps. / Interactions are an integral part of ecosystems. Whether at the finest levels, such as proteins, or the broadest, such as meta-communities, it is possible to group them into networks and study their structure. This made it possible to demonstrate that certain structures can be observed at different levels, such as nested networks, for example. In addition, interaction networks have the property of not being fixed in time and space, which gives them a major advantage for the study of evolution. They can thus serve as a support for the study of the mechanisms involved in the evolutionary processes. However, there is not yet a methodology that has achieved consensus on the use of networks and their analysis at different organizational scales. This thesis is based on the hypothesis that networks, by virtue of their properties, are relevant to consider in order to understand evolution at different organizational scales, and offer the possibility of making links between each of them. The network-based approach, combined with the use of theoretical models, would therefore be a powerful methodological tool in expanding knowledge about the processes underlying evolution. The thesis which follows consists of six chapters whose content is as follows. It begins with an introductory chapter to the concepts of interest, in particular on evolution and coevolution. The second chapter is an introduction to the use of networks in ecology, followed by the third chapter which performs a non-exhaustive review of the methodologies developed around interaction networks. The following chapters are in a way a practical application of these methods at different levels of organization. The fourth chapter returns to an aborted stage of this doctorate which will nevertheless be used to construct the model of the following chapter. The fifth chapter focuses on coevolution and its follow-up through the interaction networks between bacteria and their viruses. Finally, the sixth chapter tracks the evolution of communities thanks to the structure of phylogenetic trees and the structure of interaction networks over time.

<p dir="ltr">Infectious diseases are caused by a multitude of organisms, ranging from viruses to bacteria, from parasites to fungi, and can be passed directly or indirectly from one person to another. Further, they continue to be a leading cause of death, especially in low-resource countries, thereby emphasizing the need for continued investigation. Understanding transmission of such diseases is vital as management or prevention of outbreaks through detection, reporting, isolation, and case management are ever-evolving. One way by which scientists can study infectious diseases is through a combination of epidemiological, genomic, and evolutionary biology approaches. This doctoral research occurred precisely at this interface, spanning across the fields of genomics, molecular biology, and epidemiology, as applied to the study of infectious disease dynamics of two separate pathogen systems (protozoan and virus).</p><p dir="ltr">The first half of this research (Chapters 1 + 2) involved the implementation of SARS-CoV-2 genomic sequencing and surveillance at Purdue University. Through this investigation in a university setting (Chapter 1), this work identified relevant variants of concern in hundreds of newly sequenced viral genomes and compared variant temporal trends with other similar university settings using publicly available data. Further phylodynamic analysis of Gamma (P.1) genomes from campus revealed multiple introductions into the Purdue community, predominantly from states within the United States. A second study (Chapter 2) assessed the transmission of variants over the course of an entire academic year from 2021-2022 in Purdue’s highly vaccinated community. This research described the rapid transition from Delta to Omicron variants and investigated variant introduction events into the campus. This comprehensive analysis showed that robust surveillance programs coupled with viral genomic sequencing and phylogenetic analysis can provide critical insights into SARS-CoV-2 spread and can help inform mitigation strategies for future pandemics.</p><p dir="ltr">The latter half of this body of research (Chapters 3 + 4) focused on malaria, which is a disease caused by <i>Plasmodium </i>species<i> </i>parasites and transmitted to humans through the bites of infected mosquitoes. The first investigation explored diagnostic accuracy metrics across a malaria transmission gradient in Zambia through a comparison of the diagnostic performance of Rapid Diagnostic Tests (RDT) and Light Microscopy (LM) with photo-induced electron transfer polymerase chain reaction (PET-PCR) as the gold standard using 2018 Malaria Indicator Survey (MIS) data. Results suggested that RDTs and LM both performed well across a range of transmission intensities, but low parasitaemia infections can affect accuracy. This suggests that more sensitive tools should be utilized to identify the last cases as Zambia moves towards malaria elimination. In addition to diagnostic metrics, preventing disease is also crucial for infectious diseases, and vaccines present one mechanism by which this can be done. Research to develop a malaria vaccine with sustained high efficacy has spanned decade. However, the process has proven to be challenging, with several vaccine candidates having advanced to early-stage trials, but only a few demonstrating sustained efficacy in clinical testing. The goal of the last investigation (Chapter 4) was to shed light on the diversity of <i>Plasmodium falciparum </i>antigens which could be considered when developing future malaria vaccines. Results of evolutionary and genomic analyses of Whole-Genome-Sequences from Zambia and other countries in Africa suggest that conserved merozoite antigens and/or transmission-blocking antigens should be prioritized when developing future malaria vaccines.</p>

Les perturbations auxquelles font face les communautés écologiques, du fait des activités humaines, sont à l'origine de changements profonds dans ces communautés. Nombreuses caractéristiques des espèces sont altérées, de leur physiologie à leur occurrence même. Ces changements se répercutent sur la composition, la diversité et la structure des communautés, puisque les espèces n'interagissent pas tout le temps de la même manière en fonction des conditions. Prévoir le devenir de ces communautés émergentes, et des fonctions qu'elles soutiennent est un défi central de l'écologie et de nos sociétés. Différents cadres conceptuels ont été utilisés pour relever ce défi, basés sur différents mécanismes écologiques, et ont divergé en plusieurs domaines. D'un côté, l'analyse des chaînes trophiques utilise la consommation pour expliquer les effets de la diversité verticale (le nombre de niveaux trophiques) sur le fonctionnement, et de l'autre côté, les analyses biodiversité-fonctionnement lient compétition et effets de la diversité horizontale (la diversité au sein des niveaux trophiques isolés). Chacun de ces domaines a produit des résultats clés pour comprendre les conséquences fonctionnelles des changements de composition et diversité des communautés écologiques. Cependant, ils sont chacun basés sur différentes simplifications fortes des communautés. L'hypothèse qui sous-tend cette thèse est que la réconciliation en un même cadre de travail des résultats fondamentaux de ces champs conceptuels divergents, ainsi que des effets des changements de structure de la biodiversité, est une étape clé pour pouvoir améliorer notre compréhension du fonctionnement de communautés écologiques en changement. L'essor récent des méthodes d'analyse des réseaux trophiques, et des modèles permettant de simuler le fonctionnement de ces réseaux trophiques offre un cadre idéal pour cette réconciliation. En effet, les réseaux trophiques cartographient les échanges de matière entre toutes les espèces d'une communauté, permettant la mise en place d'interactions variées. Ils reflètent mieux la réalité complexe des communautés que les chaînes trophiques ou leurs niveaux trophiques isolés en intégrant notamment compétition et consommation. Un modèle ressource-consommateur bioénergétique classique, développé par Yodzis et Innes (1992), permet d'en simuler le fonctionnement, en intégrant des mécanismes et taux testés empiriquement. Au-delà d'utiliser ces outils, cette thèse se concentre aussi sur leur évaluation. Après un premier chapitre d'introduction, le second chapitre propose une plateforme ouverte, commune, solidement testée et efficace pour l'utilisation du modèle bioénergétique, permettant ainsi une synthèse plus rapide et aisée des résultats. Le troisième chapitre est une revue du corpus méthodologique d'analyse des réseaux trophiques, proposant une gamme de méthodes robustes et informatives, et soulignant leur domaine d'application et leurs limites. Enfin le quatrième chapitre met ce cadre méthodologique à l'épreuve. Dans ce chapitre, nous montrons l'existence d'une relation entre la complexité de la structure du réseau trophique des communautés et leur régime de fonctionnement, se traduisant par la réalisation de différentes prédictions issues de l'analyse des chaînes trophiques ou des analyses diversité-fonctionnement. Cette mise en évidence des conditions structurelles pour la réalisation de différentes prédictions nous permet de mieux comprendre quels mécanismes écologiques prédominent selon différentes conditions, dirigeant l'effet de la diversité sur le fonctionnement. / Human-driven disturbances are causing profound changes in ecological communities, as many characteristics of species are altered, from their physiology to their very occurrence. These changes affect the composition, diversity and structure of communities, since species do not always interact in the same way under different conditions. Predicting the fate of these emerging communities, and the functions they support, is a central challenge for ecology and our societies. Diverging conceptual frameworks have been used to address this challenge, based on different ecological mechanisms. On the one hand, food chain analysis uses consumption to explain the effects of vertical diversity (the number of trophic levels) on functioning, and on the other hand, biodiversity-functioning analyses link competition and the effects of horizontal diversity (diversity within isolated trophic levels). Each of these domains has produced key results for understanding the functional consequences of changes in the composition and diversity of ecological communities. However, they are each based on different strong simplifications of communities. The hypothesis underlying this thesis is that reconciling the fundamental results of these divergent conceptual fields, as well as the effects of changes in the structure of biodiversity, into a single framework is a key step towards improving our understanding of the functioning of changing ecological communities. The recent development of food web analysis and of models to simulate food webs functioning provides an ideal framework for this reconciliation. Food webs map the exchange of matter between all species in a community, allowing for a variety of interactions to take place. They better reflect the complex reality of communities than food chains or their isolated trophic levels, notably by integrating competition and consumption. A classical consumer-resource bioenergetic model developed by Yodzis and Innes (1992) specifically makes it possible to realistically simulate their functioning, using empirically tested mechanisms and rates. Beyond using these tools, this thesis focuses on their evaluation and implementation. After a first, introductory chapter, the second chapter proposes an open, common, well-tested and efficient platform for the use of the bioenergetic model, allowing a faster and easier synthesis of the results. The third chapter is a review of the methodological corpus for ecological networks analysis, outlining a range of robust and informative methods, and highlighting their scope and limitations. Finally, the fourth chapter puts this methodological framework to the test. In this chapter, we show the existence of a relationship between the complexity of communities' food-web structure and functioning regime, resulting in the realization of different predictions from food chain analysis or diversity-functioning analyses. This demonstration of the structural conditions for the realization of different predictions allows us to better understand which ecological mechanisms predominate under different conditions, directing the effect of diversity on functioning.