This thesis develops an Artificial Intelligence (AI) approach intended for accurate patient stratification and precise diagnostics/prognostics in clinical and preclinical applications. The rapid advance in high throughput technologies and bioinformatics tools is still far from linking precisely the genome-phenotype interactions with the biological mechanisms that underlie pathophysiological conditions. In practice, the incomplete knowledge on individual heterogeneity in complex diseases keeps forcing clinicians to settle for surrogate endpoints and therapies based on a generic one-size-fits-all approach. The working hypothesis is that AI can add new tools to elaborate and integrate together in new features or structures the rich information now available from high-throughput omics and bioimaging data, and that such re- structured information can be applied through predictive models for the precision medicine paradigm, thus favoring the creation of safer tailored treatments for specific patient subgroups. The computational techniques in this thesis are based on the combination of dimensionality reduction methods with Deep Learning (DL) architectures to learn meaningful transformations between the input and the predictive endpoint space. The rationale is that such transformations can introduce intermediate spaces offering more succinct representations, where data from different sources are summarized. The research goal was attacked at increasing levels of complexity, starting from single input modalities (omics and bioimaging of different types and scales), to their multimodal integration. The approach also deals with the key challenges for machine learning (ML) on biomedical data, i.e. reproducibility, stability, and interpretability of the models. Along this path, the thesis contribution is thus the development of a set of specialized AI models and a core framework of three tools of general applicability: i. A Data Analysis Plan (DAP) for model selection and evaluation of classifiers on omics and imaging data to avoid selection bias. ii. The histolab Python package that standardizes the reproducible pre-processing of Whole Slide Images (WSIs), supported by automated testing and easily integrable in DL pipelines for Digital Pathology. iii. Unsupervised and dimensionality reduction techniques based on the UMAP and TDA frameworks for patient subtyping. The framework has been successfully applied on public as well as original data in precision oncology and predictive toxicology. In the clinical setting, this thesis has developed1: 1. (DAPPER) A deep learning framework for evaluation of predictive models in Digital Pathology that controls for selection bias through properly designed data partitioning schemes. 2. (RADLER) A unified deep learning framework that combines radiomics fea- tures and imaging on PET-CT images for prognostic biomarker development in head and neck squamous cell carcinoma. The mixed deep learning/radiomics approach is more accurate than using only one feature type. 3. An ML framework for automated quantification tumor infiltrating lymphocytes (TILs) in onco-immunology, validated on original pathology Neuroblastoma data of the Bambino Gesu’ Children’s Hospital, with high agreement with trained pathologists. The network-based INF pipeline, which applies machine learning models over the combination of multiple omics layers, also providing compact biomarker signatures. INF was validated on three TCGA oncogenomic datasets. In the preclinical setting the framework has been applied for: 1. Deep and machine learning algorithms to predict DILI status from gene expression (GE) data derived from cancer cell lines on the CMap Drug Safety dataset. 2. (ML4TOX) Deep Learning and Support Vector Machine models to predict potential endocrine disruption of environmental chemicals on the CERAPP dataset. 3. (PathologAI) A deep learning pipeline combining generative and convolutional models for preclinical digital pathology. Developed as an internal project within the FDA/NCTR AIRForce initiative and applied to predict necrosis on images from the TG-GATEs project, PathologAI aims to improve accuracy and reduce labor in the identification of lesions in predictive toxicology. Furthermore, GE microarray data were integrated with histology features in a unified multi-modal scheme combining imaging and omics data. The solutions were developed in collaboration with domain experts and considered promising for application.

<p dir="ltr">Crohn’s disease and ulcerative colitis are chronic inflammatory bowel diseases (IBD) with a rising global prevalence, influenced by clinical and demographics factors. The pathogenesis of IBD involves complex interactions between gut microbiome dysbiosis, epithelial cell barrier disruption, and immune hyperactivity, which are poorly understood. This necessitates the development of novel approaches to integrate and model multiple clinical and molecular data modalities from patients, animal models, and <i>in-vitro</i> systems to discover effective biomarkers for disease progression and drug response. As sequencing technologies advance, the amount of molecular and compositional data from paired measurements of host and microbiome systems is exploding. While it is become routine to generate such rich, deep datasets, tools for their interpretation lag behind. Here, I present a computational framework for integrative modeling of microbiome multi-omics data titled: Latent Interacting Variable Effects (LIVE) modeling. LIVE combines various types of microbiome multi-omics data using single-omic latent variables (LV) into a structured meta-model to determine the most predictive combinations of multi-omics features predicting an outcome, patient group, or phenotype. I implemented and tested LIVE using publicly available metagenomic and metabolomics data set from Crohn’s Disease (CD) and ulcerative colitis (UC) status patients in the PRISM and LLDeep cohorts. The findings show that LIVE reduced the number of features interactions from the original datasets for CD to tractable numbers and facilitated prioritization of biological associations between microbes, metabolites, enzymes, clinical variables, and a disease status outcome. LIVE modeling makes a distinct and complementary contribution to the current methods to integrate microbiome data to predict IBD status because of its flexibility to adapt to different types of microbiome multi-omics data, scalability for large and small cohort studies via reliance on latent variables and dimensionality reduction, and the intuitive interpretability of the meta-model integrating -omic data types.</p><p dir="ltr">A novel application of LIVE modeling framework was associated with sex-based differences in UC. Men are 20% more likely to develop this condition and 60% more likely to progress to colitis-associated cancer compared to women. A possible explanation for this observation is differences in estrogen signaling among men and women in which estrogen signaling may be protective against UC. Extracting causal insights into how gut microbes and metabolites regulate host estrogen receptor β (ERβ) signaling can facilitate the study of the gut microbiome’s effects on ERβ’s protective role against UC. Supervised LIVE models<b> </b>ERβ signaling using high-dimensional gut microbiome data by controlling clinical covariates such as: sex and disease status. LIVE models predicted an inhibitory effect on ER-UP and ER-DOWN signaling activities by pairs of gut microbiome features, generating a novel of catalog of metabolites, microbial species and their interactions, capable of modulating ER. Two strongly positively correlated gut microbiome features: <i>Ruminoccocus gnavus</i><i> </i>with acesulfame and <i>Eubacterium rectale</i><i> </i>with 4-Methylcatechol were prioritized as suppressors ER-UP and ER-DOWN signaling activities. An <i>in-vitro</i> experimental validation roadmap is proposed to study the synergistic relationships between metabolites and microbiota suppressors of ERβ signaling in the context of UC. Two i<i>n-vitro</i> systems, HT-29 female colon cancer cell and female epithelial gut organoids are described to evaluate the effect of gut microbiome on ERβ signaling. A detailed experimentation is described per each system including the selection of doses, treatments, metrics, potential interpretations and limitations. This experimental roadmap attempts to compare experimental conditions to study the inhibitory effects of gut microbiome on ERβ signaling and how it could elevate or reduce the risk of developing UC. The intuitive interpretability of the meta-model integrating -omic data types in conjunction with the presented experimental validation roadmap aim to transform an artificial intelligence-generated big data hypothesis into testable experimental predictions.</p>

La qualité des lipides des préparations pour nourrissons est primordiale, notamment en termes d’acides gras polyinsaturés (AGPI) comme l’acide arachidonique (ARA) et docosahexaénoïque (DHA). Ces derniers pourraient favoriser le développement neurosensoriel de l’enfant. Nous avons ainsi évalué 4 standards commerciaux contenant des lipides végétaux ou laitiers et supplémentés ou non en ARA/DHA, sur le développement neurosensoriel au travers d’un modèle d’allaitement artificiel « pups in the cups ». En TEP-cs, nous observons que la supplémentation en ARA/DHA permet de normaliser le fonctionnement cérébral.L’exploration des lipides tissulaires indique des différences en DHA particulièrement bas avec l’allaitement en lipides végétaux purs. Nous proposons un algorithme de prédiction du DHA cérébrale et oculaire via les profils en acides gras érythrocytaires. Dans ces tissus un tiers des espèces à DHA sont affectées et corrélées à l’activité cérébrale. Les neuromédiateurs issus de l’AL, ARA, DHA par la voie LOX sont impactés ainsi que la distribution spatiale en DHA en IMS. Les autres données omiques soulignaient l’impact des interactions fond lipidique x ajout DHA/ARA (transcriptomique) ou fond lipidique (métabolomique) sur la régulation du métabolisme cérébral impactant le métabolisme neuronal et le métabolisme cérébral du microbiote probablement via l’axe de signalisation intestin-cerveau. Nous identifions alors un métagénome sensible à l’ajout DHA/ARA corrélé à la fonction cérébrale. Enfin, des modifications épigénétiques (méthylation du génome et miARN) touchant le groupe FC suggèrent potentiellement un impact à long terme. / The quality of lipids in infant formula is essential, especially in terms of polyunsaturated fatty acids (PUFAs) such as arachidonic acid (ARA) and docosahexaenoic acid (DHA). These could promote the neurosensory development of the child. We thus evaluated 4 commercial standards containing plant or dairy lipids and supplemented or not with ARA / DHA, on the neurosensory development through an artificially feeding model "pups in the cups". In PET-cs, we observe that the supplementation in ARA / DHA makes it possible to normalize the cerebral functioning. The exploration of tissue lipids indicates differences in DHA which are particularly low with pure plant lipids intake. We propose an algorithm for predicting cerebral and ocular DHA via erythrocyte fatty acids profiles. In these tissues one-third of the DHA species are affected and correlated with brain activity. The neuromediators resulting from AL, ARA, DHA by the LOX pathway are impacted as well as the spatial distribution of DHA in IMS imaging. Other omics data underlined the impact of lipid background x combination DHA / ARA (transcriptomics) or lipid background (metabolomics) on the regulation of cerebral metabolism impacting neuronal metabolism and brain metabolism of the microbiota probably through the signalling of gut-brain axis. We then identify a metagenome sensitive to the addition of DHA / ARA correlated to brain function. Finally, epigenetic modifications (methylation of the genome and miRNA) affecting the FC group potentially suggest a long-term impact.

The utilization of mathematical tools within biology and medicine has traditionally been less widespread compared to other hard sciences, such as physics and chemistry. However, an increased need for tools such as data processing, bioinformatics, statistics, and mathematical modeling, have emerged due to advancements during the last decades. These advancements are partly due to the development of high-throughput experimental procedures and techniques, which produce ever increasing amounts of data. For all aspects of biology and medicine, these data reveal a high level of inter-connectivity between components, which operate on many levels of control, and with multiple feedbacks both between and within each level of control. However, the availability of these large-scale data is not synonymous to a detailed mechanistic understanding of the underlying system. Rather, a mechanistic understanding is gained first when we construct a hypothesis, and test its predictions experimentally. Identifying interesting predictions that are quantitative in nature, generally requires mathematical modeling. This, in turn, requires that the studied system can be formulated into a mathematical model, such as a series of ordinary differential equations, where different hypotheses can be expressed as precise mathematical expressions that influence the output of the model. Within specific sub-domains of biology, the utilization of mathematical models have had a long tradition, such as the modeling done on electrophysiology by Hodgkin and Huxley in the 1950s. However, it is only in recent years, with the arrival of the field known as systems biology that mathematical modeling has become more commonplace. The somewhat slow adaptation of mathematical modeling in biology is partly due to historical differences in training and terminology, as well as in a lack of awareness of showcases illustrating how modeling can make a difference, or even be required, for a correct analysis of the experimental data. In this work, I provide such showcases by demonstrating the universality and applicability of mathematical modeling and hypothesis testing in three disparate biological systems. In Paper II, we demonstrate how mathematical modeling is necessary for the correct interpretation and analysis of dominant negative inhibition data in insulin signaling in primary human adipocytes. In Paper III, we use modeling to determine transport rates across the nuclear membrane in yeast cells, and we show how this technique is superior to traditional curve-fitting methods. We also demonstrate the issue of population heterogeneity and the need to account for individual differences between cells and the population at large. In Paper IV, we use mathematical modeling to reject three hypotheses concerning the phenomenon of facilitation in pyramidal nerve cells in rats and mice. We also show how one surviving hypothesis can explain all data and adequately describe independent validation data. Finally, in Paper I, we develop a method for model selection and discrimination using parametric bootstrapping and the combination of several different empirical distributions of traditional statistical tests. We show how the empirical log-likelihood ratio test is the best combination of two tests and how this can be used, not only for model selection, but also for model discrimination. In conclusion, mathematical modeling is a valuable tool for analyzing data and testing biological hypotheses, regardless of the underlying biological system. Further development of modeling methods and applications are therefore important since these will in all likelihood play a crucial role in all future aspects of biology and medicine, especially in dealing with the burden of increasing amounts of data that is made available with new experimental techniques. / Användandet av matematiska verktyg har inom biologi och medicin traditionellt sett varit mindre utbredd jämfört med andra ämnen inom naturvetenskapen, såsom fysik och kemi. Ett ökat behov av verktyg som databehandling, bioinformatik, statistik och matematisk modellering har trätt fram tack vare framsteg under de senaste decennierna. Dessa framsteg är delvis ett resultat av utvecklingen av storskaliga datainsamlingstekniker. Inom alla områden av biologi och medicin så har dessa data avslöjat en hög nivå av interkonnektivitet mellan komponenter, verksamma på många kontrollnivåer och med flera återkopplingar både mellan och inom varje nivå av kontroll. Tillgång till storskaliga data är emellertid inte synonymt med en detaljerad mekanistisk förståelse för det underliggande systemet. Snarare uppnås en mekanisk förståelse först när vi bygger en hypotes vars prediktioner vi kan testa experimentellt. Att identifiera intressanta prediktioner som är av kvantitativ natur, kräver generellt sett matematisk modellering. Detta kräver i sin tur att det studerade systemet kan formuleras till en matematisk modell, såsom en serie ordinära differentialekvationer, där olika hypoteser kan uttryckas som precisa matematiska uttryck som påverkar modellens output. Inom vissa delområden av biologin har utnyttjandet av matematiska modeller haft en lång tradition, såsom den modellering gjord inom elektrofysiologi av Hodgkin och Huxley på 1950‑talet. Det är emellertid just på senare år, med ankomsten av fältet systembiologi, som matematisk modellering har blivit ett vanligt inslag. Den något långsamma adapteringen av matematisk modellering inom biologi är bl.a. grundad i historiska skillnader i träning och terminologi, samt brist på medvetenhet om exempel som illustrerar hur modellering kan göra skillnad och faktiskt ofta är ett krav för en korrekt analys av experimentella data. I detta arbete tillhandahåller jag sådana exempel och demonstrerar den matematiska modelleringens och hypotestestningens allmängiltighet och tillämpbarhet i tre olika biologiska system. I Arbete II visar vi hur matematisk modellering är nödvändig för en korrekt tolkning och analys av dominant-negativ-inhiberingsdata vid insulinsignalering i primära humana adipocyter. I Arbete III använder vi modellering för att bestämma transporthastigheter över cellkärnmembranet i jästceller, och vi visar hur denna teknik är överlägsen traditionella kurvpassningsmetoder. Vi demonstrerar också frågan om populationsheterogenitet och behovet av att ta hänsyn till individuella skillnader mellan celler och befolkningen som helhet. I Arbete IV använder vi matematisk modellering för att förkasta tre hypoteser om hur fenomenet facilitering uppstår i pyramidala nervceller hos råttor och möss. Vi visar också hur en överlevande hypotes kan beskriva all data, inklusive oberoende valideringsdata. Slutligen utvecklar vi i Arbete I en metod för modellselektion och modelldiskriminering med hjälp av parametrisk ”bootstrapping” samt kombinationen av olika empiriska fördelningar av traditionella statistiska tester. Vi visar hur det empiriska ”log-likelihood-ratio-testet” är den bästa kombinationen av två tester och hur testet är applicerbart, inte bara för modellselektion, utan också för modelldiskriminering. Sammanfattningsvis är matematisk modellering ett värdefullt verktyg för att analysera data och testa biologiska hypoteser, oavsett underliggande biologiskt system. Vidare utveckling av modelleringsmetoder och tillämpningar är därför viktigt eftersom dessa sannolikt kommer att spela en avgörande roll i framtiden för biologi och medicin, särskilt när det gäller att hantera belastningen från ökande datamängder som blir tillgänglig med nya experimentella tekniker.

L’augmentation des rendements est un enjeu majeur chez les céréales. Dans cet objectif, il est nécessaire de maintenir la qualité du grain de blé, qui est principalement déterminée par sa teneur et sa composition en protéines de réserve. En effet, une forte relation négative existe entre le rendement et la teneur en protéines. Par ailleurs, la qualité du grain est fortement influencée par la disponibilité en azote et en soufre dans le sol. La limitation des apports d’intrants azotés à la culture et la carence en soufre récemment observée dans les sols représentent ainsi des difficultés supplémentaires pour maitriser cette qualité. Une meilleure connaissance des mécanismes moléculaires impliqués dans le contrôle du développement du grain et la mise en place de ses réserves protéiques en réponse à la nutrition azotée et soufrée est donc primordiale. L’objectif de cette thèse a ainsi été d’apporter de nouveaux éléments à la compréhension de ces processus de régulation, aujourd’hui peu connus. Pour cela, les approches -omiques sont apparues comme une stratégie de choix pour identifier les acteurs moléculaires mis en jeu. Le protéome nucléaire a été une cible importante dans les travaux menés. L’étude de ces protéines nucléaires a révélé certains régulateurs transcriptionnels qui pourraient être impliqués dans le contrôle de la mise en place des réserves du grain. Dans une approche combinant des données de protéomique, transcriptomique et métabolomique, une vision intégrative de la réponse du grain à la nutrition azotée et soufrée a été obtenue. L’importance d’un apport de soufre dans le contrôle de la balance azote/soufre du grain, déterminante pour la composition du grain en protéines de réserve, a été clairement vérifiée. Parmi les changements observés au niveau du métabolisme cellulaire, certains des gènes affectés par la modification de cette balance pourraient orchestrer l’ajustement de la composition du grain face à des situations de carences nutritionnelles. Ces nouvelles connaissances devraient permettre de mieux maitriser la qualité du grain de blé dans un contexte d’agriculture durable. / Improving the yield potential of cereals represents a major challenge. In this context, wheat grain quality has to be maintained. Indeed, grain quality is mainly determined by the content and the composition of storage proteins, but there is a strongly negative correlation between yield and grain protein concentration. In addition, grain quality is strongly influenced by the availability of nitrogen and sulfur in soils. Nowadays, the limitation of nitrogen inputs, and also the sulfur deficiency recently observed in soils represent major difficulties to control the quality. Therefore, understanding of molecular mechanisms controlling grain development and accumulation of storage proteins in response to nitrogen and sulfur supply is a major issue. The objective of this thesis was to create knowledge on the comprehension of these regulatory mechanisms. For this purpose, the best strategy to identify molecular actors involved in these processes consisted of -omics approaches. In our studies, the nuclear proteome was an important target. Among these proteins, we revealed some transcriptional regulators likely to be involved in the control of the accumulation of grain storage compounds. Using an approach combining proteomic, transcriptomic and metabolomic data, the characterization of the integrative grain response to the nitrogen and sulfur supply was obtained. Besides, our studies clearly confirmed the major influence of sulfur in the control of the nitrogen/sulfur balance that determines the grain storage protein composition. Among the changes observed in the cell metabolism, some genes were disturbed by the modification of this balance. Thus these genes could coordinate the adjustment of grain composition in response to nutritional deficiencies. These new results contribute in facing the challenge of maintaining wheat grain quality with sustainable agriculture.

Depuis plusieurs années, des projets d'intégration de données omiques se sont développés, notamment avec objectif de participer à la description fine de caractères complexes d'intérêt socio-économique. Dans ce contexte, l'objectif de cette thèse est de combiner différentes données omiques hétérogènes afin de mieux décrire et comprendre le dernier tiers de gestation chez le porc, période influençant la mortinatalité porcine. Durant cette thèse, nous avons identifié les bases moléculaires et cellulaires sous-jacentes de la fin de gestation, en particulier au niveau du muscle squelettique. Ce tissu est en effet déterminant à la naissance car impliqué dans l'efficacité de plusieurs fonctions physiologiques comme la thermorégulation et la capacité à se déplacer. Au niveau du plan expérimental, les tissus analysés proviennent de foetus prélevés à 90 et 110 jours de gestation (naissance à 114 jours), issus de deux lignées extrêmes pour la mortalité à la naissance, Large White et Meishan, et des deux croisements réciproques. Au travers l'application de plusieurs études statistiques et computationnelles (analyses multidimensionnelles, inférence de réseaux, clustering et intégration de données), nous avons montré l'existence de mécanismes biologiques régulant la maturité musculaire chez les porcelets, mais également chez d'autres espèces d'intérêt agronomique (bovin et mouton). Quelques gènes et protéines ont été identifiées comme étant fortement liées à la mise en place du métabolisme énergétique musculaire durant le dernier tiers de gestation. Les porcelets ayant une immaturité du métabolisme musculaire seraient sujets à un plus fort risque de mortalité à la naissance. Un second volet de cette thèse concerne l'imputation de données manquantes (tout un groupe de variables pour un individu) dans les méthodes d'analyses multidimensionnelles, comme l'analyse factorielle multiple (AFM) (ou multiple factor analysis (MFA)). Dans notre contexte, l'AFM fut particulièrement intéressante pour l'intégration de données d'un ensemble d'individus sur différents tissus (deux ou plus). Afin de conserver ces individus manquants pour tout un groupe de variables, nous avons développé une méthode, appelée MI-MFA (multiple imputation - MFA), permettant l'estimation des composantes de l'AFM pour ces individus manquants. / Over the last decades, some omics data integration studies have been developed to participate in the detailed description of complex traits with socio-economic interests. In this context, the aim of the thesis is to combine different heterogeneous omics data to better describe and understand the last third of gestation in pigs, period influencing the piglet mortality at birth. In the thesis, we better defined the molecular and cellular basis underlying the end of gestation, with a focus on the skeletal muscle. This tissue is specially involved in the efficiency of several physiological functions, such as thermoregulation and motor functions. According to the experimental design, tissues were collected at two days of gestation (90 or 110 days of gestation) from four fetal genotypes. These genotypes consisted in two extreme breeds for mortality at birth (Meishan and Large White) and two reciprocal crosses. Through statistical and computational analyses (descriptive analyses, network inference, clustering and biological data integration), we highlighted some biological mechanisms regulating the maturation process in pigs, but also in other livestock species (cattle and sheep). Some genes and proteins were identified as being highly involved in the muscle energy metabolism. Piglets with a muscular metabolism immaturity would be associated with a higher risk of mortality at birth. A second aspect of the thesis was the imputation of missing individual row values in the multidimensional statistical method framework, such as the multiple factor analysis (MFA). In our context, MFA was particularly interesting in integrating data coming from the same individuals on different tissues (two or more). To avoid missing individual row values, we developed a method, called MI-MFA (multiple imputation - MFA), allowing the estimation of the MFA components for these missing individuals.

La teneur et la composition en protéines de réserve du grain de blé tendre (Triticum aestivum L.) sont les principaux déterminants de sa valeur d’usage et de sa qualité nutritionnelle. La composition en protéines de réserve du grain est déterminée par la teneur en assimilâts azotés et soufrés par grain via des lois d’échelle qui pourraient être les propriétés émergentes de réseaux de régulation. Plusieurs facteurs de transcription intervenant dans cette régulation ont été mis en évidence, mais les voies et mécanismes impliqués sont encore très peu connus. Le constat est identique en ce qui concerne l’impact de la nutrition azotée et soufrée sur ce réseau de régulation. Le développement des outils de génomique fonctionnelle et de bioinformatique permet aujourd’hui d’aborder ces régulations de manière globale via une approche systémique mettant en relation plusieurs niveaux de régulation. L’objectif du travail présenté est d’explorer les réseaux de régulation –omiques impliqués dans le contrôle de l’accumulation des protéines de réserve dans le grain de blé tendre et leur réponse à l’approvisionnement en azote et en soufre. Une approche d’inférence de réseaux basée sur la découverte de règles a été étendue, implémentée sous la forme d’une plateforme web. L’utilisation de cette plateforme a permis de définir des sémantiques multiples afin d’inférer dans un cadre global, des règles possédant différentes significations biologiques. Des facteurs de transcription spécifiques de certains organes et certaines phases de développement ont été mis en évidence et un intérêt particulier a été apporté à leur position dans les réseaux de règles inférés, notamment en relation avec les protéines de réserve. Les travaux initiés dans cette thèse ouvrent un champ d’investigation innovant pour l’identification de nouvelles cibles de sélection variétale pour l’amélioration de la valeur technologique et de la qualité nutritionnelle du blé. Ils devraient ainsi permettre de mieux maîtriser la composition en protéines de réserve et ainsi produire des blés adaptés à des utilisations ciblées ou carencé en certaines fractions protéiques impliquées dans des phénomènes d’allergénicité et d’intolérance du gluten, ce dans un contexte d’agriculture durable et plus économe en intrants. / Grain storage protein content and composition are the main determinants of bread wheat (Triticum aestivum L.) end-use value. Scaling laws governing grain protein composition according to grain nitrogen and sulfur content could be the outcome of a finely tuned regulation network. Although it was demonstrated that the main regulation of grain storage proteins accumulation occurs at the transcriptomic level in cereals, knowledge of the underlying molecular mechanisms is elusive. Moreover, the effects of nitrogen and sulfur on these mechanisms are unknown. The issue of skyrocketing data generation in research projects is addressed by developing high-throughput bioinformatics approaches. Extracting knowledge on from such massive amounts of data is therefore an important challenge. The work presented herein aims at elucidating regulatory networks involved in grain storage protein synthesis and their response to nitrogen and sulfur supply using a rule discovery approach. This approach was extended, implemented in the form of a web-oriented platform dedicated to the inference and analysis of regulatory networks from qualitative and quantitative –omics data. This platform allowed us to define different semantics in a comprehensive framework; each semantic having its own biological meaning, thus providing us with global informative networks. Spatiotemporal specificity of transcription factors expression was observed and particular attention was paid to their relationship with grain storage proteins in the inferred networks. The work initiated here opens up a field of innovative investigation to identify new targets for plant breeding and for an improved end-use value and nutritional quality of wheat in the context of inputs limitation. Further analyses should enhance the understanding of the control of grain protein composition and allow providing wheat adapted to specific uses or deficient in protein fractions responsible for gluten allergenicity and intolerance.

La médecine de précision (MP) est un nouveau paradigme qui révolutionne la pratique médicale actuelle et remodèle complètement la médecine de demain. La MP aspire à placer le patient au centre du parcours de soins en y intégrant les données médicales et biologiques individuelles tout en tenant compte de la grande diversité interindividuelle. La prédiction des états pathologiques chez les patients nécessite une compréhension dynamique et systémique. Les erreurs innées du métabolisme (EIM) sont des troubles génétiques résultant de défauts dans une voie biochimique donnée en raison de la déficience d'une enzyme, de son cofacteur ou d’un transporteur. Les EIM ne sont plus considérées comme des maladies monogéniques mais tendent à être plus complexes et multifactorielles. Le profil métabolomique permet le dépistage d’une pathologie, la recherche de biomarqueurs et l’exploration des voies métaboliques mises en jeu. Dans ce travail de thèse, nous avons utilisé l’approche métabolomique qui est particulièrement pertinente pour les EIM compte tenu de leur physiopathologie de base qui est étroitement liée au métabolisme. Ce travail a permis la mise en place d’une méthodologie métabolomique non ciblée basée sur une stratégie analytique multidimensionnelle comportant la spectrométrie de masse à haute résolution couplée à la chromatographie liquide ultra-haute performance et la mobilité ionique. La mise en place de la méthodologie de prétraitement, d’analyse et d’exploitation des données générées avec des outils de design expérimental et d’analyses multivariées ont été aussi établies. Enfin, cette approche a été appliquée pour l’exploration des EIM avec les mucopolysaccharidoses comme preuve de concept. Les résultats obtenus suggèrent un remodelage majeur du métabolisme des acides aminés dans la mucopolysaccharidose de type I. En résumé, la métabolomique pourrait être un outil complémentaire pertinent en appui à l’approche génomique dans l’exploration des EIM. / The new field of precision medicine is revolutionizing current medical practice and reshaping future medicine. Precision medicine intends to put the patient as the central driver of healthcare by broadening biological knowledge and acknowledging the great diversity of individuals. The prediction of physiological and pathological states in patients requires a dynamic and systemic understanding of these interactions. Inborn errors of metabolism (IEM) are genetic disorders resulting from defects in a given biochemical pathway due to the deficiency of an enzyme, its cofactor or a transporter. IEM are no longer considered to be monogenic diseases, which adds another layer of complexity to their characterization and diagnosis. To meet this need for faster screening, the metabolic profile can be a promising candidate given its ability in disease screening, biomarker discovery and metabolic pathway investigation. In this thesis, we used a metabolomic approach which is particularly relevant for IEM given their basic pathophysiology that is tightly related to metabolism. This thesis allowed the implementation of an untargeted metabolomic methodology based on a multidimensional analytical strategy including high-resolution mass spectrometry coupled with ultra-high-performance liquid chromatography and ion mobility. This work also set a methodology for preprocessing, analysis and interpretation of the generated data using experimental design and multivariate data analysis. Finally, the strategy is applied to the exploration of IEM with mucopolysaccharidoses as a proof of concept. The results suggest a major remodeling of the amino acid metabolisms in mucopolysaccharidosis type I. In summary, metabolomic is a relevant complementary tool to support the genomic approach in the functional investigations and diagnosis of IEM.

Despite major advances in treatment, pediatric cancers in the 5-16 age group remain the most common cause of disease death, and one out of eight children with cancer will not survive. Among children that do survive, some 60% suffer from late effects such as cancer recurrence and increased risk of obesity. This paper will provide a broad overview of pediatric oncology in the context of systems medicine. Systems medicine utilizes an integrative approach that relies on patient information gained from omics technology. A major goal of a systems medicine is to provide personalized medicine that optimizes positive outcomes while minimizing deleterious short and long-term sideeffects. There is an ever increasing development of effective cancer drugs, but a major challenge lies in picking the most effective drug for a particular patient. As detailed below, high-throughput omics technology holds the promise of solving this problem. Omics includes genomics, epigenomics, and proteomics. System medicine integrates omics information and provides detailed insights into disease mechanisms which can then inform the optimal treatment strategy.

Tesis por compendio / [ES] En los últimos años, el aprendizaje profundo (DL) se ha convertido en una de las principales áreas de la inteligencia artificial (IA), impulsado principalmente por el avance en la capacidad de procesamiento. Los algoritmos basados en DL han logrado resultados asombrosos en la comprensión y manipulación de diversos tipos de datos, incluyendo imágenes, señales de habla y texto. La revolución digital del sector sanitario ha permitido la generación de nuevas bases de datos, lo que ha facilitado la implementación de modelos de DL bajo el paradigma de aprendizaje supervisado. La incorporación de estos métodos promete mejorar y automatizar la detección y el diagnóstico de enfermedades, permitiendo pronosticar su evolución y facilitar la aplicación de intervenciones clínicas de manera más efectiva. Una de las principales limitaciones de la aplicación de algoritmos de DL supervisados es la necesidad de grandes bases de datos anotadas por expertos, lo que supone una barrera importante en el ámbito médico. Para superar este problema, se está abriendo un nuevo campo de desarrollo de estrategias de aprendizaje no supervisado o débilmente supervisado que utilizan los datos disponibles no anotados o débilmente anotados. Estos enfoques permiten aprovechar al máximo los datos existentes y superar las limitaciones de la dependencia de anotaciones precisas. Para poner de manifiesto que el aprendizaje débilmente supervisado puede ofrecer soluciones óptimas, esta tesis se ha enfocado en el desarrollado de diferentes paradigmas que permiten entrenar modelos con bases de datos débilmente anotadas o anotadas por médicos no expertos. En este sentido, se han utilizado dos modalidades de datos ampliamente empleadas en la literatura para estudiar diversos tipos de cáncer y enfermedades inflamatorias: datos ómicos e imágenes histológicas. En el estudio sobre datos ómicos, se han desarrollado métodos basados en deep clustering que permiten lidiar con las altas dimensiones inherentes a este tipo de datos, desarrollando un modelo predictivo sin la necesidad de anotaciones. Al comparar el método propuesto con otros métodos de clustering presentes en la literatura, se ha observado una mejora en los resultados obtenidos. En cuanto a los estudios con imagen histológica, en esta tesis se ha abordado la detección de diferentes enfermedades, incluyendo cáncer de piel (melanoma spitzoide y neoplasias de células fusocelulares) y colitis ulcerosa. En este contexto, se ha empleado el paradigma de multiple instance learning (MIL) como línea base en todos los marcos desarrollados para hacer frente al gran tamaño de las imágenes histológicas. Además, se han implementado diversas metodologías de aprendizaje, adaptadas a los problemas específicos que se abordan. Para la detección de melanoma spitzoide, se ha utilizado un enfoque de aprendizaje inductivo que requiere un menor volumen de anotaciones. Para abordar el diagnóstico de colitis ulcerosa, que implica la identificación de neutrófilos como biomarcadores, se ha utilizado un enfoque de aprendizaje restrictivo. Con este método, el coste de anotación se ha reducido significativamente al tiempo que se han conseguido mejoras sustanciales en los resultados obtenidos. Finalmente, considerando el limitado número de expertos en el campo de las neoplasias de células fusiformes, se ha diseñado y validado un novedoso protocolo de anotación para anotaciones no expertas. En este contexto, se han desarrollado modelos de aprendizaje profundo que trabajan con la incertidumbre asociada a dichas anotaciones. En conclusión, esta tesis ha desarrollado técnicas de vanguardia para abordar el reto de la necesidad de anotaciones precisas que requiere el sector médico. A partir de datos débilmente anotados o anotados por no expertos, se han propuesto novedosos paradigmas y metodologías basados en deep learning para abordar la detección y diagnóstico de enfermedades utilizando datos ómicos e imágenes histológicas. / [CA] En els últims anys, l'aprenentatge profund (DL) s'ha convertit en una de les principals àrees de la intel·ligència artificial (IA), impulsat principalment per l'avanç en la capacitat de processament. Els algorismes basats en DL han aconseguit resultats sorprenents en la comprensió i manipulació de diversos tipus de dades, incloent-hi imatges, senyals de parla i text. La revolució digital del sector sanitari ha permés la generació de noves bases de dades, la qual cosa ha facilitat la implementació de models de DL sota el paradigma d'aprenentatge supervisat. La incorporació d'aquests mètodes promet millorar i automatitzar la detecció i el diagnòstic de malalties, permetent pronosticar la seua evolució i facilitar l'aplicació d'intervencions clíniques de manera més efectiva. Una de les principals limitacions de l'aplicació d'algorismes de DL supervisats és la necessitat de grans bases de dades anotades per experts, la qual cosa suposa una barrera important en l'àmbit mèdic. Per a superar aquest problema, s'està obrint un nou camp de desenvolupament d'estratègies d'aprenentatge no supervisat o feblement supervisat que utilitzen les dades disponibles no anotades o feblement anotats. Aquests enfocaments permeten aprofitar al màxim les dades existents i superar les limitacions de la dependència d'anotacions precises. Per a posar de manifest que l'aprenentatge feblement supervisat pot oferir solucions òptimes, aquesta tesi s'ha enfocat en el desenvolupat de diferents paradigmes que permeten entrenar models amb bases de dades feblement anotades o anotades per metges no experts. En aquest sentit, s'han utilitzat dues modalitats de dades àmpliament emprades en la literatura per a estudiar diversos tipus de càncer i malalties inflamatòries: dades ómicos i imatges histològiques. En l'estudi sobre dades ómicos, s'han desenvolupat mètodes basats en deep clustering que permeten bregar amb les altes dimensions inherents a aquesta mena de dades, desenvolupant un model predictiu sense la necessitat d'anotacions. En comparar el mètode proposat amb altres mètodes de clustering presents en la literatura, s'ha observat una millora en els resultats obtinguts. Quant als estudis amb imatge histològica, en aquesta tesi s'ha abordat la detecció de diferents malalties, incloent-hi càncer de pell (melanoma spitzoide i neoplàsies de cèl·lules fusocelulares) i colitis ulcerosa. En aquest context, s'ha emprat el paradigma de multiple instance learning (MIL) com a línia base en tots els marcs desenvolupats per a fer front a la gran grandària de les imatges histològiques. A més, s'han implementat diverses metodologies d'aprenentatge, adaptades als problemes específics que s'aborden. Per a la detecció de melanoma spitzoide, s'ha utilitzat un enfocament d'aprenentatge inductiu que requereix un menor volum d'anotacions. Per a abordar el diagnòstic de colitis ulcerosa, que implica la identificació de neutròfils com biomarcadores, s'ha utilitzat un enfocament d'aprenentatge restrictiu. Amb aquest mètode, el cost d'anotació s'ha reduït significativament al mateix temps que s'han aconseguit millores substancials en els resultats obtinguts. Finalment, considerant el limitat nombre d'experts en el camp de les neoplàsies de cèl·lules fusiformes, s'ha dissenyat i validat un nou protocol d'anotació per a anotacions no expertes. En aquest context, s'han desenvolupat models d'aprenentatge profund que treballen amb la incertesa associada a aquestes anotacions. En conclusió, aquesta tesi ha desenvolupat tècniques d'avantguarda per a abordar el repte de la necessitat d'anotacions precises que requereix el sector mèdic. A partir de dades feblement anotades o anotats per no experts, s'han proposat nous paradigmes i metodologies basats en deep learning per a abordar la detecció i diagnòstic de malalties utilitzant dades *ómicos i imatges histològiques. Aquestes innovacions poden millorar l'eficàcia i l'automatització en la detecció precoç i el seguiment de malalties. / [EN] In recent years, deep learning (DL) has become one of the main areas of artificial intelligence (AI), driven mainly by the advancement in processing power. DL-based algorithms have achieved amazing results in understanding and manipulating various types of data, including images, speech signals and text. The digital revolution in the healthcare sector has enabled the generation of new databases, facilitating the implementation of DL models under the supervised learning paradigm. Incorporating these methods promises to improve and automate the detection and diagnosis of diseases, allowing the prediction of their evolution and facilitating the application of clinical interventions with higher efficacy. One of the main limitations in the application of supervised DL algorithms is the need for large databases annotated by experts, which is a major barrier in the medical field. To overcome this problem, a new field of developing unsupervised or weakly supervised learning strategies using the available unannotated or weakly annotated data is opening up. These approaches make the best use of existing data and overcome the limitations of reliance on precise annotations. To demonstrate that weakly supervised learning can offer optimal solutions, this thesis has focused on developing different paradigms that allow training models with weakly annotated or non-expert annotated databases. In this regard, two data modalities widely used in the literature to study various types of cancer and inflammatory diseases have been used: omics data and histological images. In the study on omics data, methods based on deep clustering have been developed to deal with the high dimensions inherent to this type of data, developing a predictive model without requiring annotations. In comparison, the results of the proposed method outperform other existing clustering methods. Regarding histological imaging studies, the detection of different diseases has been addressed in this thesis, including skin cancer (spitzoid melanoma and spindle cell neoplasms) and ulcerative colitis. In this context, the multiple instance learning (MIL) paradigm has been employed as the baseline in all developed frameworks to deal with the large size of histological images. Furthermore, diverse learning methodologies have been implemented, tailored to the specific problems being addressed. For the detection of spitzoid melanoma, an inductive learning approach has been used, which requires a smaller volume of annotations. To address the diagnosis of ulcerative colitis, which involves the identification of neutrophils as biomarkers, a constraint learning approach has been utilized. With this method, the annotation cost has been significantly reduced while achieving substantial improvements in the obtained results. Finally, considering the limited number of experts in the field of spindle cell neoplasms, a novel annotation protocol for non-experts has been designed and validated. In this context, deep learning models that work with the uncertainty associated with such annotations have been developed. In conclusion, this thesis has developed cutting-edge techniques to address the medical sector's challenge of precise data annotation. Using weakly annotated or non-expert annotated data, novel paradigms and methodologies based on deep learning have been proposed to tackle disease detection and diagnosis in omics data and histological images. These innovations can improve effectiveness and automation in early disease detection and monitoring. / The work of Rocío del Amor to carry out this research and to elaborate this dissertation has been supported by the Spanish Ministry of Universities under the FPU grant FPU20/05263. / Amor Del Amor, MRD. (2023). Deep Learning Strategies for Overcoming Diagnosis Challenges with Limited Annotations [Tesis doctoral]. Universitat Politècnica de València. https://doi.org/10.4995/Thesis/10251/200227 / Compendio