Tesi sul tema "Bio-statistics"

Nous présentons dans cette thèse plusieurs travaux de statistiques bayésiennes appliquées à la génétique des populations. La génétique des populations a pour but d'expliquer les variations génétiques au sein d'une espèce, et d'inférer les processus ayant conduits à ces variations. Pour cela, des données génétiques massives sont utilisées et il y a un besoin grandissant de méthodes statistiques pour traiter ces données. Le travail de cette thèse s'inscrit dans cet effort de modélisation statistique pour répondre aux enjeux de la génétique des populations, et de la biologie de l'évolution. Nous nous intéressons tout particulièrement à la détection de traces d'adaptation locale dans les génomes, et à l'inférence des variations spatiales non stationnaires.Un modèle d'analyse factorielle bayésien est proposé pour détecter les traces d'adaptation locale. Nous comparons notre approche aux méthodes existantes, et démontrons qu'elle permet d'obtenir un plus faible taux de fausses découvertes. Nous présentons également un modèle bayésien basé sur des processus gaussiens pour caractériser les variations génétiques spatiales dans l'aire de répartition d'une espèce. Les performances de ces méthodes sont démontrées sur différents exemples issus de simulations ou de données. Plusieurs logiciels open source qui implémentent ces méthodes ont été développés pendant la thèse
In this thesis we present several works related to Bayesian statistics in population genetics. Population genetics aims at explaining genetic variation within natural species, and infer the different processes that lead to current genetic variation. Large scale genomic datasets are produced, and there is an increasing need of statistical methods to extract information from these datasets. My thesis work is part of this statistical modeling effort to answer to evolutionary biology and population genetic questions. We are interested in detecting footprints of local adaptation without, and infering non-stationary patterns of spatial variation. A Bayesian factor model is used to detect genes involved in local adaptation. We compare our factor model to existing methods, and show that it can reduce the false discovery rate. We also present a Bayesian model based on Gaussian processes to caracterize spatial genetic variations within species. The performances of these methods are tested on simulations and real datasets. Several open source software are available online

Les avancées des nouvelles technologies de séquençage ont ouvert l'accès aux données d'expression dynamique des gènes à l'échelle du génome. Les approches ensemblistes classiques, habituellement utilisées en biologie, ne permettent pas la compréhension des mécanismes moléculaires complexes sous-jacents. Par conséquent, le développement de méthodes analytiques permettant d'appréhender de manière plus satisfaisante ce type de données représente un défi majeur pour la biologie contemporaine. Cependant, les coûts techniques et expérimentaux associés aux données de transcriptomiques limitent la dimension des jeux de données réels et, par conséquent, leurs méthodes d'analyse. Au cours de ma thèse, à l'interface entre les mathématiques appliquées et la biologie végétale, j'ai travaillé sur la mise en place d'une méthode d'inférence de réseaux de régulations dynamiques adaptée à un jeu de données réelles et originales décrivant l'effet de stimulations sonores répétées sur l'expression des gènes d'Arabidopsis thaliana. J'ai ainsi proposé une méthode de classification adaptée aux séries temporelles très courtes qui regroupe les gènes par variations temporelles, permettant d'ajuster la dimension des données à l'inférence de réseau. La comparaison de cette méthode aux méthodes classiques a permis de montrer qu'elle était la plus adaptée aux séries temporelles très courtes avec un pas de temps irrégulier. Pour l'inférence de réseau dynamique, j'ai proposé un modèle qui prend en compte la variabilité intra-classe et qui intègre un terme constant décrivant explicitement la stimulation externe du système. L'évaluation de ces méthodes de classification et d'inférence a été effectuée sur des données de séries temporelles simulées et réelles, ce qui a permis d'établir la bonne qualité des performance en terme de précision, de rappel et d'erreur de prédiction. L'implémentation de ces méthodes a permis d'étudier le priming de la réponse immunitaire d'Arabidopsis thaliana par des ondes sonores répétées. Nous avons montré l'existence de la formation d'une mémoire transcriptionnelle associée aux stimulations qui fait passer la plante d'un état naïf à un état primé et plus résistant en 3 jours. Cet état résistant, entretenu d'une part par les stimulations et d'autre part par des cascades de facteurs de transcription, augmente la résistance immunitaire de la plante en déclenchant l'expression de gènes de résistance chez la plante saine, en diversifiant le nombre de gènes participant à la réponse immunitaire et en intensifiant l'expression de nombreux gènes de résistance. L'inférence du réseau décrivant la mémoire transcriptionnelle associée aux stimulations sonores répétées nous a permis d'identifier les propriétés qu'elle confère à la plante. Ces prédictions, validées expérimentalement, ont montré par exemple que l'augmentation de la cadence entre stimulations ne permettait pas d'obtenir un gain de résistance plus conséquent et que la mémoire transcriptionnelle ne dure que 1.5 jours après la dernière stimulation
Advancements in new sequencing technologies have paved the way for accessing dynamic gene expression data on a genome-wide scale. Classical ensemble approaches traditionally used in biology fall short of comprehending the underlying the complex molecular mechanisms. Consequently, developing analytical methods to understand further such data poses a significant challenge for current biology. However, the technical and experimental costs associated with transcriptomic data severely limit the dimension of real datasets and their analytical methods. Throughout my thesis, at the intersection of applied mathematics and plant biology, I focused on implementing an inference method for dynamic regulatory networks tailored to a real and original dataset describing the effect of repeated acoustic stimulations on genes expressions of Arabidopsis thaliana. I proposed a clustering method adapted to very-short time series that groups genes based on temporal variations, adjusting the data dimension for network inference. The comparison of this method with classical methods showed that it was the most suitable for very-short time series with irregular time points. For the network inference, I proposed a model that takes into account intra-class variability and integrates a constant term explicitly describing the external stimulation of the system. The evaluation of these classification and inference methods was conducted on simulated and real-time series data, which established their high performance in terms of accuracy, recall, and prediction error. The implementation of these methods to study the priming of the immune response of Arabidopsis thaliana through repeated sound waves. We demonstrated the formation of a transcriptional memory associated with stimulations, transitioning the plant from a naïve state to a primed and more resistant state within 3 days. This resistant state, maintained by stimulations and transcription factor cascades, enhances the plant's immune resistance by triggering the expression of resistance genes in healthy plants, diversifying the number of genes involved in the immune response, and intensifying the expression of numerous resistance genes. The inference of the network describing the transcriptional memory associated with repeated sound stimulations allowed us to identify the properties conferred to plants. Experimentally validated predictions showed that increasing the frequency between stimulations does not result in a more significant resistance gain, and the transcriptional memory lasts only 1.5 days after the last stimulation

La valutazione del potenziale carcinogeno delle sostanze chimiche si basa, a livello regolatorio, sul saggio di cancerogenesi sui roditori (OECD, TG 451). La loro diffusa applicabilità è però messa in discussione a causa delle lunghe tempistiche e dei costi associati alla loro esecuzione, e dal potenzialmente limitato valore predittivo verso l'uomo. I Cell Transformation Assays (CTA) sono tra i metodi in vitro più avanzati e standardizzati a livello regolatorio. I CTA riproducono passaggi chiave della trasformazione in vivo; si basano sulla formazione, in seguito al trattamento con un sospetto carcinogeno, di colonie trasformate (o foci), classificate al microscopio da un esperto sulla base di caratteristiche morfologiche standard. I CTA sono impiegati sia per la valutazione del rischio sia nell'ambito della ricerca; tuttavia, per consentirne un maggior utilizzo, alcuni miglioramenti sono auspicabili. Nel presente lavoro sono stati sviluppati due approcci allo scopo di potenziare i saggi CTA. Il primo ha lo scopo di ottimizzare la fase di riconoscimento dei foci trasformati sulla base di caratteristiche morfologiche, spesso affetta da soggettività. Questo obiettivo è stato raggiunto sviluppando algoritmi e metodi statistici per definire in modo quantitativo le caratteristiche morfologiche dei foci. Il secondo approccio ha come obiettivo il miglioramento della comprensione dei meccanismi coinvolti nella trasformazione in vitro, attraverso la caratterizzazione biochimica e molecolare di cellule isolate da foci e durante fasi iniziali della trasformazione. E' stato acquisito un database di immagini di foci ottenuti tramite saggi CTA condotti da EURL ECVAM (JRC), nell'ambito dello studio di prevalidazione, e da ARPA-ER. Sono stati sviluppati: 1) un algoritmo di segmentazione capace di isolare la regione del focus dal monostrato circostante; 2) descrittori statistici delle immagini dei foci, allo scopo di riassumere le caratteristiche di dimensione, crescita multistrato, invasività e grado di eterogeneità dei foci, e spindle-shape delle cellule. I descrittori statistici così definiti sono stati impiegati per costruire modelli di classificazione dei foci, allo scopo di fornire supporto alla classificazione operata dall'esperto. Inoltre è stato possibile studiare l’effetto esercitato dalla concentrazione di due carcinogeni sulla morfologia dei foci trasformati, come catturata dai descrittori statistici introdotti. Allo scopo di migliorare la comprensione dei meccanismi di cancerogenesi in vitro, sono state valutate diverse fasi temporali del processo. Campioni provenienti dalle fasi iniziali del processo di trasformazione e da foci trasformati sono stati analizzati tramite tecniche di biochimica e trascrittomica. Il saggio CTA è un metodo utilizzato ampiamente in accademia per studiare i meccanismi di azione della cancerogenesi chimica; abbiamo sfruttato questa configurazione sperimentale per analizzare i processi coinvolti nella cancerogenesi in vitro indotta da cadmio, cancerogeno umano i cui meccanismi di azione sono ancora non completamente noti. Abbiamo dimostrato che nelle fasi iniziali a seguito di trattamento con cadmio sono coinvolti processi riguardanti l'omeostasi degli ioni Cd2+, Zn2+ e Ca2+, alterazioni del citoscheletro e della segnalazione cellulare. A partire da questi processi di difesa, possono essere intraprese vie non univoche per completare la trasformazione: a fronte dello stesso stimolo possono originarsi foci caratterizzati da diversi fenotipi, che sottendono a diversi profili biochimici/molecolari. L'inclusione nei saggi CTA di una valutazione oggettiva delle caratteristiche morfologiche e della corrispondenza tra il fenotipo dei foci trasformati e le caratteristiche molecolari alla base del processo di trasformazione, rappresenta un significativo avanzamento nella valutazione della cancerogenesi in vitro, componente di un più generale approccio integrato alla valutazione del rischio.
The evaluation of the carcinogenic potential to humans relies at regulatory level on the two-year rodent bioassays (OECD TG451), which are extremely costly in terms of time and animals used, and whose predictive value towards humans has been questioned. The Cell Transformation Assays (CTAs) are the most advanced in vitro methods to identify the chemical carcinogenicity potential, in terms of standardization and validation, and reproduce key stages of in vivo transformation. The endpoint is the formation of transformed colonies (or foci) upon treatment with a carcinogen, which are visually scored by stereomicroscopy, using defined morphological features. These assays offer several advantages in comparison to the in vivo bioassays in rodents, and are used by industry and academia as screening methods for carcinogenicity testing and as a tool for mechanistic studies. Even though OECD Guidance Documents on CTAs have been recently published, further improvements are considered important to enhance the use of the assay. We developed two approaches aiming to: i) increase throughput and reliability of the scoring process, by developing algorithms and statistical methods designed to quantitatively characterize foci morphological features and ii) increase the understanding of in vitro transformation mechanisms, through the molecular characterization of transformed cells from foci, and from initial phases of transformation. A database of digital images of foci from CTAs performed by EURL ECVAM (JRC), in the prevalidation study, and by ARPA-ER, were acquired and foci regions were isolated from the background through an originally developed algorithm. Statistical image descriptors defining the morphological features recognized during visual scoring were developed to cover size, multilayer growth, spindle-shape, invasiveness, and degree of heterogeneity of foci. Statistical models were developed to automatically classify foci, supporting the phase of visual scoring. In addition, exploiting fitted parametric models using defined statistical descriptors, it was possible to estimate the effects of concentrations of tested carcinogens on foci morphologies. To disclose the mechanisms of in vitro transformation, it is crucial to evaluate the process through a temporal approach. Cells from initial phases of exposure to carcinogen and from transformed foci were collected to perform transcriptomic and biochemical analyses of signalling cascades. We exploited this system to study mechanisms involved in cadmium-induced transformation, hence cadmium is an established human carcinogen, but whose mechanisms of action are still not fully understood. During in vitro transformation many processes are involved and non-unique ways to the establishment of transformed cells can be covered. Indeed, we demonstrated that upon the same stimulus, foci characterised by different phenotypes can be induced, and different phenotypes correspond indeed to a specific biochemical/molecular cell clone fingerprint. This approach provides a tool for mechanistic studies and it will allow the comprehension of the links between transformed phenotype of foci and the biochemical fingerprint. An increased mechanistic understanding of in vitro transformation could support an integrated approach based on quantitative scoring of foci and molecular fingerprinting. This advancement will also meet specific recommendations of EURL ECVAM in view of future broader acceptance of these assays.

The main goal of population genetics is to understand the factors that affect genetic variation within a species. Mathematical models are used to predict the effects on genetic variation of processes such as mutation, recombination, selection, migration and population size changes, but analytical results are difficult to obtain when these processes interact and when equilibrium conditions are not met. In these situations, common in real biological systems especially when recent human activities (e.g., stocking, urbanization, overhunting) perturb natural populations, computer simulations can be very useful. A computer simulation is a virtual experiment in which a model is used to mimic the biological process on a computer to study its properties. It is an excellent tool for understanding the functioning of complex systems. Simulations are generally used to make predictions about populations, validate statistical methods, study the properties of different sampling strategies, and estimate parameters from real data. In this thesis, I applied genetic simulations to address questions intractable with other methods. First, I analyzed the effects of violating the assumption of panmixiamade by “Extende Bayesian Skyline Plot” (EBSP) method. I showed that migration can influence the inferred demographic history of a population, suggesting wrong dynamics. Second, I used genetic simulations to analyse the performance of the EBSP method in reconstructing a population decline and to compare sampling schemes with different proportions of modern and ancient DNA. I identified some properties of the sampling scheme which clearly positively affect the demographic reconstruction, providing some simple hints for planning a genetic study when both modern and ancient samples are available. Third, I familiarized with the “Approximated Bayesian Computation” methodology and I contributed to a review article presenting the main features, with pros and cons, of this approach. Fourth, I applied the ABC procedure to analyze the hybridization history within the genus Chionodraco, and to evaluate the power of ABC in this context. Realistic demographic models were defined and compared, and evidence was found that hybridization occurred only in interglacial periods. Taken together, the results presented in this thesis confirm the importance of genetic simulations in evolutionary biology. If we consider the increasing availability of simulation packages, along with the increasing speed and storage capacity of personal computers and clusters, it is easy to predict that simulations of genetic and genomic data will spread in many fields to better explore more and more realistic, and consequently complex, models.

Les approches statistiques en génétique des populations visent deux objectifs distincts qui sont la description des données et la possibilité d'inférer les processus évolutifs qui ont généré les patrons observés. Le premier chapitre de ce manuscrit décrit nos apports théoriques et méthodologiques concernant le calcul bayésien approché (Approximate Bayesian Computation) qui permet de réaliser l'objectif d'inférence des processus évolutifs. Je décris des résultats asymptotiques qui permettent de décrire des propriétés statistiques du calcul bayésien approché. Ces résultats mettent en évidence à la fois l'intérêt des méthodes dites avec ajustement qui reposent sur des équations de régression et aussi l'intérêt de réduire la dimension des descripteurs statistiques utilisés dans le calcul bayésien approché. Je présente ensuite une méthode originale de calcul bayésien approché qui permet de manière conjointe d'effectuer des ajustements et de réduire la dimension des descripteurs statistiques. Une comparaison des différentes méthodes de réduction de dimension clos le premier chapitre. Le deuxième chapitre est consacré à l'objectif de description des données et se place plus particulièrement dans un cadre spatial. Les méthodes statistiques proposées reposent sur le concept d'isolement par la distance qui est une forme particulière de l'autocorrélation spatiale où la corrélation entre individus décroit avec la distance. Une approche originale de krigeage nous permet de caractériser des patrons d'isolement par la distance non-stationnaire où la manière avec laquelle la corrélation entre individus décroit avec la distance dépend de l'espace. Une deuxième extension que nous proposons est celle d'isolement par la distance anisotrope que nous caractérisons et testons à partir d'une équation de régression. La conclusion de ce manuscrit met l'accent sur les problèmes d'interprétation des résultats statistiques, l'importance de l'échantillonnage et la nécessité de tester l'adéquation des modèles aux données. Je conclus par des perspectives qui se proposent de faire passer l'analyse statistique bayésienne à l'échelle des données massives produites en génétique.

Modélisation du coût thermodynamique des structures pseudoneuds.
Développement d'un algorithme d'accélération exacte de la dynamique de monte Carlo.
Développement et interfaçage web d'un code de repliement de molécule ARN incluant les motifs pseudoneuds. Rendu cinématique de la dynamique de repliement.
Etude statistique de la prévalence des pseudoneouds dans des séquences biologiques et aléatoires.
Vérification expérimentale du code repliement. Démonstration expérimentale de l'existence d'un super-code guidant pour le repliement natif des ARN.

Cette thèse est à l'intersection de deux domaines de recherche en plein expansion, à savoir la fouille de données et la bio-informatique. Avec l'émergence des bases de graphes au cours des dernières années, de nombreux efforts ont été consacrés à la fouille des sous-graphes fréquents. Mais le nombre de sous-graphes fréquents découverts est exponentiel, cela est due principalement à la nature combinatoire des graphes. Beaucoup de sous-graphes fréquents ne sont pas pertinents parce qu'ils sont redondants ou tout simplement inutiles pour l'utilisateur. En outre, leur nombre élevé peut nuire ou même rendre parfois irréalisable toute utilisation ultérieure. La redondance dans les sous-graphes fréquents est principalement due à la similarité structurelle et / ou sémantique, puisque la plupart des sous-graphes découverts diffèrent légèrement dans leur structures et peuvent exprimer des significations similaires ou même identiques. Dans cette thèse, nous proposons deux approches de sélection des sous-graphes représentatifs parmi les fréquents a n d'éliminer la redondance. Chacune des approches proposées s'intéresse à un type spécifique de redondance. La première approche s'adresse à la redondance sémantique où la similarité entre les sous-graphes est mesurée en fonction de la similarité entre les étiquettes de leurs nœuds, en utilisant les connaissances de domaine. La deuxième approche s'adresse à la redondance structurelle où les sous-graphes sont représentés par des descripteurs topologiques définis par l'utilisateur, et la similarité entre les sous-graphes est mesurée en fonction de la distance entre leurs descriptions topologiques respectives. Les principales données d'application de cette thèse sont les structures 3D des protéines. Ce choix repose sur des raisons biologiques et informatiques. D'un point de vue biologique, les protéines jouent un rôle crucial dans presque tous les processus biologiques. Ils sont responsables d'une variété de fonctions physiologiques. D'un point de vue informatique, nous sommes intéressés à la fouille de données complexes. Les protéines sont un exemple parfait de ces données car elles sont faites de structures complexes composées d'acides aminés interconnectés qui sont eux-mêmes composées d'atomes interconnectés. Des grandes quantités de structures protéiques sont actuellement disponibles dans les bases de données en ligne. Les structures 3D des protéines peuvent être transformées en graphes où les acides aminés représentent les nœuds du graphe et leurs connexions représentent les arêtes. Cela permet d'utiliser des techniques de fouille de graphes pour les étudier. L'importance biologique des protéines et leur complexité ont fait d'elles des données d'application appropriées pour cette thèse.

Evaluer, maintenir et restaurer les conditions écologiques des rivières nécessitent des mesures du fonctionnement de leurs écosystèmes. De par leur complexité, notre compréhension de ces systèmes est imparfaite. La prise en compte des incertitudes et variabilités liées à leur évaluation est donc indispensable à la prise de décision des gestionnaires. En analysant des données nationales (~ 1654 sites), les objectifs principaux de cette thèse étaient de (1) tester certaines hypothèses intrinsèques aux bio-indicateurs et (2) d'étudier les incertitudes de l'évaluation écologique associées à la variabilité temporelle des bio-indicateurs et à la prédiction des conditions de référence. (1) Ce travail met en évidence (i) le rôle prépondérant des facteurs environnementaux naturels dans la structuration des communautés aquatiques en comparaison des facteurs anthropiques (définis à l'échelle du bassin versant, du corridor riparien et du tronçon), (ii) les réponses contrastées des communautés aquatiques aux pressions humaines (dégradations hydro-morphologiques et de la qualité de l'eau) et (iii) plus généralement, les forts impacts des barrages et de l'altération de la qualité de l'eau sur les communautés aquatiques. (2) Une méthode Bayésienne a été développée pour estimer les incertitudes liées à la prédiction des conditions de référence d'un indice piscicole (IPR+). Les incertitudes prédictives de l'IPR+ dépendent du site considéré mais aucune tendance claire n'a été observée. Par comparaison, la variabilité temporelle de l'IPR+ est plus faible et semble augmenter avec l'intensité des perturbations anthropiques. Les résultats de ce travail confirment l'avantage d'indices multi-métriques basés sur des traits fonctionnels par rapport à ceux relatifs à la composition taxonomique. Les sensibilités différentes des macrophytes, poissons, diatomées et macro-invertébrés aux pressions humaines soulignent leur complémentarité pour l'évaluation des écosystèmes fluviaux. Néanmoins, de futures recherches sont nécessaires à une meilleure compréhension des effets d'interactions entre types de pressions et entre pressions humaines et environnement
Sensitive biological measures of ecosystem quality are needed to assess, maintain or restore the ecological conditions of rivers. Since our understanding of these complex systems is imperfect, river management requires recognizing variability and uncertainty of bio-assessment for decision-making. Based on the analysis of national data sets (~ 1654 sites), the main goals of this work were (1) to test some of the assumptions that shape bio-indicators and (2) address the temporal variability and the uncertainty associated to prediction of reference conditions.(1) This thesis highlights (i) the predominant role of physiographic factors in shaping biological communities in comparison to human pressures (defined at catchment, riparian corridor and reach scales), (ii) the differences in the responses of biological indicators to the different types of human pressures (water quality, hydrological, morphological degradations) and (iii) more generally, the greatest biological impacts of water quality alterations and impoundments. (2) A Bayesian method was developed to estimate the uncertainty associated with reference condition predictions of a fish-based bio-indicator (IPR+). IPR+ predictive uncertainty was site-dependent but showed no clear trend related to the environmental gradient. By comparison, IPR+ temporal variability was lower and sensitive to an increase of human pressure intensity. This work confirmed the advantages of multi-metric indexes based on functional metrics in comparison to compositional metrics. The different sensitivities of macrophytes, fish, diatoms and macroinvertebrates to human pressures emphasize their complementarity in assessing river ecosystems. Nevertheless, future research is needed to better understand the effects of interactions between pressures and between pressures and the environment

La découverte des anesthésiques représente un progrès majeur de la médecine, rendu possible par l'observation empirique de leurs effets. Des expériences ont révélées les neurorécepteurs comme cibles possibles des anesthésiques, des canaux localisés sur la membrane des cellules cibles aux terminaisons nerveuses. Le récepteur GLIC, un homologue bactérien du récepteur nicotinique humain, a été co-cristallisé en 2011 avec des anesthésiques généraux liés à lui. Dans cette thèse, j'utilise les simulations de dynamique moléculaire pour caractériser les interactions entre des anesthésiques généraux et différentes formes de GLIC. En 2011, le propofol et le desflurane ont été co-cristallisés dans un site de liaison intra-sous-unité localisé dans le domaine transmembranaire de GLIC. En 2013, il a été montré que le bromoforme se lie à ce site ainsi qu'à un site inter-sous-unités. Dans ce travail, je décris des simulations d'une nouvelle structure cristalline montrant un site de liaison situé dans le pore du canal. Des simulations d'innondation de GLIC par le bromoforme ont démontré l'accessibilité spontanée des sites expérimentaux dans un environnement non cristallin. Des calculs d'énergie libre mettent en évidence des différences d'énergie de liaison entre les sites et entre des mutants de GLIC. Un échantillonnage complet des poches de liaison m'a permis de détecter un autre site de liaison inter-sous-unité duquel l'accessibilité est semble modulée par un résidu en particulier. Les données accumulées au cours de ce projet fournissent une image grandissante de l'action des anesthésiques à l'échelle atomique.

Neuronal networks are today hypothesized to the basis for the computing capabilities of biological nervous systems. In the same manner, artificial neuronal systems are intensively exploited for a diversity of industrial and scientific applications. However, how information is represented and processed by these networks remains under debate, meaning that it is not clear which sets of neuronal activity features are useful for computation. In this thesis, I present a set of results that link the first-order statistics of neuronal activity with behavior, in the general context of encoding/decoding to analyse experimental data collected while non human primates performed a working memory task. Subsequently, I go beyond the first-order and show that the second-order statistics of neuronal activity in reservoir computing, a recurrent artificial network model, make up a robust candidate for information representation and transmission for the classification of multivariate inputs.
Las redes neuronales se presentan hoy, hipotéticamente, como las responsables de las capacidades computacionales de los sistemas nerviosos biológicos. De la misma manera, los sistemas neuronales artificiales son intensamente explotados en una diversidad de aplicaciones industriales y científicas. No obstante, cómo la información es representada y procesada por estas redes está aún sujeto a debate. Es decir, no está claro qué propiedades de la actividad neuronal son útiles para llevar a cabo computaciones. En esta tesis, presento un conjunto de resultados que relaciona el primer orden estadístico de la actividad neuronal con comportamiento, en el contexto general de codificación/decodificación, para analizar datos recolectados mientras primates no humanos realizaban una tarea de memoria de trabajo. Subsecuentemente, voy más allá del primer orden y muestro que las estadísticas de segundo orden en computación de reservorios, un modelo de red neuronal artificial y recurrente, constituyen un candidato robusto para la representación y transmisión de información con el fin de clasificar señales multidimensionales.

Submitted by Priscila Nascimento (pnascimento@icict.fiocruz.br) on 2012-11-12T18:43:04Z No. of bitstreams: 1 ricardo-de-carvalho.pdf: 1945198 bytes, checksum: 8586b33e6fabb5d0036c0443e9c26caa (MD5)
Made available in DSpace on 2012-11-12T18:43:04Z (GMT). No. of bitstreams: 1 ricardo-de-carvalho.pdf: 1945198 bytes, checksum: 8586b33e6fabb5d0036c0443e9c26caa (MD5) Previous issue date: 2005
Fundação Oswaldo Cruz. Instituto de Tecnologia em Imunobiológicos. Rio de Janeiro, RJ, Brasil / Fundação Oswaldo Cruz. Instituto Oswaldo Cruz Rio de Janeiro, RJ, Brasil.
Esta dissertação investiga a aplicação do Controle Estatístico de Processo (CEP) nas principais fases da produção da suspensão viral (SV) da vacina contra febre amarela produzida em Bio-Manguinhos/FIOCRUZ, Brasil. Empregada na formulação do produto final, a SV é preparada em várias etapas, utilizando a tecnologia aperfeiçoada no Brasil (egg passage), desde 1937. O sistema de CEP implantado está baseado em recursos computacionais, perfeitamente adaptados à realidadeda Instituição. Os gráficos de controle foram a principal ferramenta estatística empregada.Sete variáveis relativas às etapas da produção da SV foram estudadas: perdas de ovos no transporte; ovos não embrionados; embriões mortos na primeira incubação; perdas na inoculação; embriões mortos na segunda incubação; perdas na coleta de embriões e rendimento na coleta de embriões. Dentre as variáveis investigadas, perdas por transporte e fertilidade dos ovos não sofrem interferência direta da produção, porém são de fundamental importância para as fases seguintes e, por esse motivo, foram estudadas. Os resultados mostraram que seis variáveis, das sete analisadas, apresentaram resultados sob controle, ainda que duas delas não tenham atendido a uma das regras sensibilizantes estabelecidas: ovos não embrionados e perdas na coleta. A variável rendimento ficou caracterizada como não controlada.Em relação à capacidade, os processos de transporte de ovos e ovos não embrionados foram classificados como processos incapazes (Cpk< 1,00). Os processos com as variáveis embriões mortos na primeira incubação e perdas na inoculação foram considerados como relativamentecapazes (1,00 ≤Cpk < 1,33) e, finalmente, os processos com variáveis embriões mortos na segunda incubação e perdas na coleta foram considerados capazes (Cpk ≥1,33). Pela análise do conjunto, os resultados indicam a necessidade de ajustes, intensificação e controle contínuo dos processos que estão sendo praticados.
This work deals with the application of Statistical Control Process (SCP) to the main parts of the production of Viral Suspension (VS) ofthe yellow fever vaccine manufactured by Bio-Manguinhos/FIOCRUZ, Brazil. The VS is used in the formulation of the final product and is prepared in several steps, using a technology (egg passage), which has been being improved in Brazil since 1937. The SCP system used is based on computational resources that are well adapted to the Institution reality. Control graphs were the main statistical tools employed. Seven variables related to the productionof the VS were investigated: transporting losses of eggs; non-germinated eggs; dead embryos during the first incubation; losses during inoculation; dead embryos during the second incubation; losses during embryo collect and efficiency in collecting embryo. Among the studied variables, transporting losses and fertility of eggs do not relate directly to the production, but are of fundamental importance for the subsequent steps, and were also studied for this reason. The results showed that six out of seven analyzed variables were under control, although two of them did not attend one of the established sensitivity rules: non germinated eggs and losses in collecting. Efficiency ended up characterized as a non-controlled variable. As far as capacity is concerned, transporting of eggs and non-germinated eggs were classified as uncapable processes (Cpk < 1,00). Dead embryos during the first incubation and losses during inoculation were considered as relatively capable processes (1,00 ≤Cpk < 1,33). Finally, dead embryos during the second incubation and losses in collecting embryo were considered capable processes (Cpk ≥1,33). The results, thus, show a necessity of adjustments,intensification and continuous control of the currently used processes.

abstract: With the increased demand for genetically modified T-cells in treating hematological malignancies, the need for an optimized measurement policy within the current good manufacturing practices for better quality control has grown greatly. There are several steps involved in manufacturing gene therapy. These steps are for the autologous-type gene therapy, in chronological order, are harvesting T-cells from the patient, activation of the cells (thawing the cryogenically frozen cells after transport to manufacturing center), viral vector transduction, Chimeric Antigen Receptor (CAR) attachment during T-cell expansion, then infusion into patient. The need for improved measurement heuristics within the transduction and expansion portions of the manufacturing process has reached an all-time high because of the costly nature of manufacturing the product, the high cycle time (approximately 14-28 days from activation to infusion), and the risk for external contamination during manufacturing that negatively impacts patients post infusion (such as illness and death). The main objective of this work is to investigate and improve measurement policies on the basis of quality control in the transduction/expansion bio-manufacturing processes. More specifically, this study addresses the issue of measuring yield within the transduction/expansion phases of gene therapy. To do so, it was decided to model the process as a Markov Decision Process where the decisions being made are optimally chosen to create an overall optimal measurement policy; for a set of predefined parameters.
Dissertation/Thesis
Masters Thesis Industrial Engineering 2020

Exploited fish stocks are assessed with the help of micro analytic and macro models (Alagaraja, 1990). Catch in numbers (age specific or length specific) or catch in weight and the corresponding fishing effort expended are the main inputs to the fish population models. The quality of this input data governs the performance (predictive or interpolative) of the models and determines the relevance of management decisions inferred from the stock assessment studies. Catches usually are estimated from sampling of commercial landings. These sampling schemes are often complex and multistage in nature. In India marine fish catch statistics are collected by the Central Marine Fisheries Research Institute, Cochin through a sampling system based on the theory of sampling(Bane~i, 1971). Most of the catches are from the inshore regions and landed at about 2400 landing centres spread all along the coast line in the various 9 maritime states of India. Keeping pace with the changing pattern of the fishery the mode of collection has also undergone change periodically without any significant alterations in the basic structure of the sampling design.

Since the discovery of AIDS among the gay men in 1981 in the United States of America, it has become a major world pandemic with over 40 million individuals infected world wide. According to the Joint United Nations Programme against HIV/AIDS epidermic updates in 2012, 28.3 million individuals are living with HIV world wide, 23.5 million among them coming from sub-saharan Africa and 4.8 million individuals residing in Asia. The report showed that approximately 1.7 million individuals have died from AIDS related deaths, 34 million ± 50% know their HIV status, a total of 2:5 million individuals are newly infected, 14:8 million individuals are eligible for HIV treatment and only 8 million are on HIV treatment (Joint United Nations Programme on HIV/AIDS and health sector progress towards universal access: progress report, 2011). Numerous studies have been carried out to understand the pathogenesis and the dynamics of this deadly disease (AIDS) but, still its pathogenesis is poorly understood. More understanding of the disease is still needed so as to reduce the rate of its acquisition. Researchers have come up with statistical and mathematical models which help in understanding and predicting the progression of the disease better so as to find ways in which its acquisition can be prevented and controlled. Previous studies on HIV/AIDS have shown that, inter-individual variability plays an important role in susceptibility to HIV-1 infection, its transmission, progression and even response to antiviral therapy. Certain immuno-genetic factors (human leukocyte antigen (HLA), Interleukin-10 (IL-10) and single nucleotide polymorphisms (SNPs)) have been associated with the variability among individuals. In this dissertation we are going to reaffirm previous studies through statistical modelling and analysis that have shown that, immuno-genetic factors could play a role in susceptibility, transmission, progression and even response to antiviral therapy. This will be done using the Sinikithemba study data from the HIV Pathogenesis Programme (HPP) at Nelson Mandela Medical school, University of Kwazulu-Natal consisting of 451 HIV positive and treatment naive individuals to model how the HIV Bio-markers (viral load and CD4 count) are associated with the immuno-genetic factors using linear mixed models. We finalize the dissertation by dealing with drop-out which is a pervasive problem in longitudinal studies, regardless of how well they are designed and executed. We demonstrate the application and performance of multiple imputation (MI) in handling drop-out using a longitudinal count data from the Sinikithemba study with log viral load as the response. Our aim is to investigate the influence of drop-out on the evolution of HIV Bio-markers in a model including selected genetic factors as covariates, assuming the missing mechanism is missing at random (MAR). We later compare the results obtained from the MI method to those obtained from the incomplete dataset. From the results, we can clearly see that there is much difference in the findings obtained from the two analysis. Therefore, there is need to account for drop-out since it can lead to biased results if not accounted for.
Thesis (M.Sc.)-University of KwaZulu-Natal, Pietermaritzburg, 2013.

Developments in modern bio-imaging techniques have allowed the routine collection of a vast amount of data from various techniques. The challenges lie in how to build accurate and efficient models to draw conclusions from the data and facilitate scientific discoveries. Fortunately, recent advances in statistics, machine learning, and deep learning provide valuable tools. This thesis describes some of our efforts to build scalable Bayesian models for four bio-imaging applications: (1) Stochastic Optical Reconstruction Microscopy (STORM) Imaging, (2) particle tracking, (3) voltage smoothing, (4) detect color-labeled neurons in c elegans and assign identity to the detections.

Les séquences protéiques naturelles sont le résultat net de l’interaction entre les mécanismes de mutation, de sélection naturelle et de dérive stochastique au cours des temps évolutifs. Les modèles probabilistes d’évolution moléculaire qui tiennent compte de ces différents facteurs ont été substantiellement améliorés au cours des dernières années. En particulier, ont été proposés des modèles incorporant explicitement la structure des protéines et les interdépendances entre sites, ainsi que les outils statistiques pour évaluer la performance de ces modèles. Toutefois, en dépit des avancées significatives dans cette direction, seules des représentations très simplifiées de la structure protéique ont été utilisées jusqu’à présent. Dans ce contexte, le sujet général de cette thèse est la modélisation de la structure tridimensionnelle des protéines, en tenant compte des limitations pratiques imposées par l’utilisation de méthodes phylogénétiques très gourmandes en temps de calcul. Dans un premier temps, une méthode statistique générale est présentée, visant à optimiser les paramètres d’un potentiel statistique (qui est une pseudo-énergie mesurant la compatibilité séquence-structure). La forme fonctionnelle du potentiel est par la suite raffinée, en augmentant le niveau de détails dans la description structurale sans alourdir les coûts computationnels. Plusieurs éléments structuraux sont explorés : interactions entre pairs de résidus, accessibilité au solvant, conformation de la chaîne principale et flexibilité. Les potentiels sont ensuite inclus dans un modèle d’évolution et leur performance est évaluée en termes d’ajustement statistique à des données réelles, et contrastée avec des modèles d’évolution standards. Finalement, le nouveau modèle structurellement contraint ainsi obtenu est utilisé pour mieux comprendre les relations entre niveau d’expression des gènes et sélection et conservation de leur séquence protéique.
Protein sequences are the net result of the interplay of mutation, natural selection and stochastic variation. Probabilistic models of molecular evolution accounting for these processes have been substantially improved over the last years. In particular, models that explicitly incorporate protein structure and site interdependencies have recently been developed, as well as statistical tools for assessing their performance. Despite major advances in this direction, only simple representations of protein structure have been used so far. In this context, the main theme of this dissertation has been the modeling of three-dimensional protein structure for evolutionary studies, taking into account the limitations imposed by computationally demanding phylogenetic methods. First, a general statistical framework for optimizing the parameters of a statistical potential (an energy-like scoring system for sequence-structure compatibility) is presented. The functional form of the potential is then refined, increasing the detail of structural description without inflating computational costs. Always at the residue-level, several structural elements are investigated: pairwise distance interactions, solvent accessibility, backbone conformation and flexibility of the residues. The potentials are then included into an evolutionary model and their performance is assessed in terms of model fit, compared to standard evolutionary models. Finally, this new structurally constrained phylogenetic model is used to better understand the selective forces behind the differences in conservation found in genes of very different expression levels.