Spécialisée dans la production d’actifs végétaux, l’entreprise PAT développe une activité innovante de recherche qui consiste à produire ou à modifier par voie enzymatique, des molécules présentes naturellement dans les plantes. L’espèce Ipomoea batatas contient de nombreux dérivés d’acide caféoyl-quinique dont majoritairement du 3,5-dicaféoyl-quinique (3,5-DCQ), une molécule antioxydante très recherchée en cosmétique. Cependant, la voie de biosynthèse est assez méconnue pour pouvoir exploiter les gènes d’intérêt par des approches d’ingénierie métabolique en vue d’augmenter la production. Les travaux réalisés dans le cadre de cette thèse ont porté sur l’identification et la caractérisation fonctionnelle d’une hydroxycinnamoyl transférase (HCT) en vue d’augmenter la production de 3,5-DCQ soit en micro-organismes soit dans des plantes recombinantes. Pour réaliser ces travaux, une banque de données RNAseq a été générée permettant ainsi d'accéder à des séquences codantes. Ces séquences ont été analysées par alignement avec des séquences codant pour des hydroxycinnamoyl transférases (HCT, HQT) impliquées dans la synthèse d’un précurseur potentiel, l’acide chlorogénique. Un premier tri a été mené en utilisant des approches d’expression différentielle de gènes. La fonction des gènes a été étudiée par des approches d’expression hétérologue en système bactérien et de transgénèse végétale. La production des métabolites cibles a été analysée dans des plantes transgéniques et dans des cultures cellulaires. Sur ces plantes, nous avons réalisé des tests de résistance à des pathogènes fongiques. Nous avons identifié une HCT qui partage 85% d’identité avec une HCT de café impliquée dans la synthèse de 3,5-DCQ. Cette activité a pu être démontrée in vitro pour l’HCT d’ipomée. De plus, l’expression du gène codant pour l’HCT conduit à une surproduction de 3,5-DCQ dans les cultures cellulaires de tabacs exprimant les HCT. Cette molécule inhibe la croissance de Botrytis cinerea et de Phytophthora parasitica. Compte tenu des teneurs en 3,5-DCQ très faibles par rapport à la plante d’origine, ces résultats suggèrent l’implication d’autres gènes dans cette voie de biosynthèse. L’activité antifongique du 3,5-DCQ pourrait être exploitée pour des applications agrochimiques / Specialized in the production of plant actives, the company PAT develops an innovate research activity that consists of producing or modifying by enzymatic pathway, molecules naturally present in plants. The species Ipomoea batatas contains numerous caffeoylquinic acid derivatives, predominantly 3,5-dicaffeoylquinic acid (3,5-DCQ), an antioxydant molecule arousing interest in cosmetics. However, biosynthesis pathway of this molecule is poorly established in order to exploit the genes of interest by metabolic engineering approaches to increase the production. The work realized in the frame of this PhD concerns the identification and functional characterization of a hydroxycinnamoyl transferase (HCT) in order to increase the production of 3,5-DCQ either in microorganisms or in recombinant plants. To perform this work, an RNAseq databank was generated allowing to access to coding sequences. These sequences were analysed by alignment of sequences encoding for hydroxycinnamoyl transferases (HCT, HQT) involved in the biosynthesis of a potential precursor, chlorogenic acid. A first screening was performed by utilizing an approach of differential expression of genes. The function of genes was studied by heterologous expression in bacterial systems and by plant transgenesis approaches. The production of target metabolites was analyzed in transgenic plants and cell cultures. On these plants, we conducted tests of resistance to fungal pathogens. We identified an HCT that shares 85% of identity with a HCT isolated from coffee previously characterized in 3,5-DCQ biosynthesis. This activity was shown in vitro for HCT of Ipomee. Moreover, the expression of target gene led to an overproduction of 3,5-DCQ in cell cultures of tobacco expressing HCT. This molecule inhibits growth of Botrytis cinerea and Phytophthora parasitica. Giving that amounts of 3,5-DCQ are very low compared to the plant of origin, these results suggest the involvement of other genes in this biosynthesis pathway. Antifungal activity of 3,5-DCQ could be exploited for agrochimic applications

Addressing issues of high complexity, such as societal or environmental transformations, require collaboration among a multitude of actors. Therefore, inter-organisational collaborations has evolved for approaching such large-scale problems. During the last decade, the orgasational theory of ecosystems has emerged further. Knowledge ecosystems is one evolved area that has limited empirical research. To evaluate how programs can be organised to sustain value to the actors, and provide further empirical research of ecosystems, this thesis explores the value creation and value capture mechanisms of a knowledge ecosystem. The thesis analyses a case study of a neutral collaborative platform, CLOSER, in Sweden. A conceptual framework, based on the open systems perspective, network theory, and the theoretical definition of a knowledge ecosystem, is used to analyse the empirical results. The value creation and value capture mechanisms for knowledge ecosystems in this empirical case study are identified as communication channels, projects and business initiatives. The findings from this study confirm the literature surrounding knowledge ecosystems and their value mechanisms. It is also elaborated that the incentives of individual participants in the ecosystem must be taken into consideration to enhance value creation and value capture, even in knowledge ecosystems. From a system level perspective, there also needs to be proper building mechanisms in place to handle the deliverance of the knowledge outside of the internal network of the ecosystem. This to ensure nothing is lost, or lessened. The notion that business ecosystems grow from knowledge ecosystems is also adressed. However, future research is still needed to provide empirical evidence of the theory. / Samhälls- eller miljötransformationer kräver samarbete mellan en mängd olika aktörer. Organiseringsfromer som bygger på interorganisatoriskt samarbete har därför utvecklats för att tillsammans kunna ta fram lösningar på sådana storskaliga problem. Ett område inom organisationsteori som har vuxit under det senaste decenniet är ekosystem, där kunskapsekosystem är ett koncept som under senare dit har blivit mer prevalent inom teoretisk forskning. Däremot saknas det underlag inom forskningen på uhr kunskapsekosystem fungerar i praktiken. Denna avhandling tillhandahåller ny empirisk forsknning inom området genom en empirisk studie på hur värde kan skapas och tas till vara på i ett kunskapsekosystem. Det görs genom att analysera hur program kan organiseras inom ett kunskapsekosystem för att upprätthålla värde för aktörerna. Avhandlingen analyserar en fallstudie av en neutral samarbetsplattform CLOSER, i Sverige. Ett konceptuellt ramverk, baserat på ett systemperspektiv, nätverksteori och den teoretiska defenitionen av ett kunskapsekosystem, används för att analysera de empiriska resultaten.  Mekanismerna för värdeskapande och värdefångande inom kunskapsekosystem i denna empiriska fallstudie identifieras som kommunikationskanaler, project och affärsincitament. Resultaten bekräftar delvis det som står i befintlig litteratur kring värdeskapande mekanismer inom kunskapsekosystem, men framhäver även att det inte räcker att fokusera på de gemensamma målen i ett storskaligt samarbete bland en mängd olika aktörer. De individuella målen som aktörerna har måste också beaktas, inklusive deras affärsmodeller, för att se till att hela ekosystemet arbetar i samma riktning. Ur ett systemnivåperspektiv måste det också finnas ordentliga uppbyggnadsmekanismer för att förmedla kusnkap utanför systemets interna nätverk för att säkerställa att värdefull kunskap inte går till spillo. Uppfattningen att affärsverksamhetsekosystem växer fram ur kunskapekosystem diskuteras också. Framtida forskning behövs dock för ayy ge empiriska bevis för den teorin.

Magister Public Health - MPH / High HIV incidence among young women is a global public health concern with the potential for a huge impact on their lives. Several interventions have been identified to help reduce HIV incidence among young women, including HIV Counselling and Testing (HCT). Lack of knowledge of HIV status is a major barrier to HIV prevention, care and treatment efforts. Therefore, HCT uptake among young women needs to be up-scaled. Early detection and treatment can help cushion the impact of AIDS on the lives of young women. There are various factors that negatively influence young women’s HCT uptake, including psychosocial characteristics. Studies have linked self-esteem to HCT and high levels of HIV risk behaviour practices among young women. The aim of this study was to assess the relationship between self-esteem and HCT uptake among young women (16 to 24 years) in South Africa using data from the 2012, Third National HIV Communication Survey (NCS). This was a quantitative secondary data analysis of cross-sectional data from the 2012 NCS. The participants of this study were young women (n =1922) from all provinces in South Africa. The independent variable was self-esteem and the dependent variable was ever testing for HIV and testing in the last 12 months, with a number of socio-demographic characteristics as covariates. Data was analysed using STATA statistical software (version 13.0, STATA Corp., College Station Texas, USA). Socio-demographic characteristics were described through descriptive statistics. The relationship between independent variables and HIV testing was analysed with Chi-squared tests of association. Logistic regression models were used to examine the relationship between self-esteem and HCT uptake when controlling for confounders.

Submitted by Kamila Costa (kamilavasconceloscosta@gmail.com) on 2015-07-23T20:59:50Z No. of bitstreams: 1 Dissertação - Ellen Suzany Pereira Aranha.pdf: 5525688 bytes, checksum: 2e52510a8984ee46894c1f0f78cdff51 (MD5) / Approved for entry into archive by Divisão de Documentação/BC Biblioteca Central (ddbc@ufam.edu.br) on 2015-07-24T19:10:26Z (GMT) No. of bitstreams: 1 Dissertação - Ellen Suzany Pereira Aranha.pdf: 5525688 bytes, checksum: 2e52510a8984ee46894c1f0f78cdff51 (MD5) / Approved for entry into archive by Divisão de Documentação/BC Biblioteca Central (ddbc@ufam.edu.br) on 2015-07-24T19:22:21Z (GMT) No. of bitstreams: 1 Dissertação - Ellen Suzany Pereira Aranha.pdf: 5525688 bytes, checksum: 2e52510a8984ee46894c1f0f78cdff51 (MD5) / Made available in DSpace on 2015-07-24T19:22:21Z (GMT). No. of bitstreams: 1 Dissertação - Ellen Suzany Pereira Aranha.pdf: 5525688 bytes, checksum: 2e52510a8984ee46894c1f0f78cdff51 (MD5) Previous issue date: 2014-04-15 / CNPq - Conselho Nacional de Desenvolvimento Científico e Tecnológico / Cancer is one of the most causes of death in the world. The research of natural products is strategy to develop anticancer agents. The aim this research was to identify chemical components, to investigate the anticancer potential in vitro and genotoxicity of Eugenia genus´s essential oil. For this, were used the cytotoxicity assay of Alamar blue and the hemolytic potential in erythrocytes of mice, as initial screening for selection of samples with cytotoxic potential, from the determination of the samples IC50. After the samples selection, we attemped to evaluate anticancer effect of the essential oil and its genotoxicity. Nine essential oils were tested, in a single concentration (50 μg/mL) which only two- Eugenia cuspidifolia (1) e Eugenia tapacumensis (3)- were selected for later tests. The IC50, in 72 hours, obtained to these samples varied between 4,69 to 24,35 μg/mL and 6,22 to 26,17 μg/mL, respectively, among tumor cell lines. For non-tumor cell line, in 72 hours, the values for the IC50 were 25,51 μg/mL to E. cuspidifolia and 36,12 μg/mL to E. tapacumensis besides they didn´t cause hemolysis to mice erythrocytes. The species E. cuspidifolia and E. tapacumensis were more active to colorectal line (HCT 116) and ovary (ES-2), and the HCT 116 line was chosen to be used for later tests, because it´s one of the most frequent kind of cancer in diagnostics and in deaths worldwide at a world level . The chemical constitution of E. cuspidifolia and E. tapacumensis essential oils was investigated through gas chromatography. A total 26 constituents were identified for both samples. The majority components were caryophyllene oxide, α-copaene, hepoxid of humulene II and cis-nerolidol. The essential oils in concentrations 7,5, 15 and 30 μg/mL, were tested in the clonogenic assay and they significantly reduced the number of colonies (p<0,05), in the 7,5 and 15 μg/mL concentrations, when compared with DMSO (0,2%). In the wound-healing assay, 24 and 48 hours, the essential oils reduced the migration in vitro (p<0,05) only on the concentration 30 μg/mL, at 48h. The inhibitor potential of metalloproteinase MMP-2 and MMP-9 was evaluated through of zymography method. In the results, the better inhibition effect was from the E. tapacumensis essential oil, that reduced the enzymatic activity (p<0,05) in the concentrations 15 and 30 μg/mL, in both treatments times (24 and 48h). Genotoxicity was evaluated by comet assay, using two versions, alkaline pH, which detects any damage caused to DNA, and neutral pH, specific to double-stranded breaks in DNA. Through damage index analisis, the results in alkaline pH assay were similar to the neutral pH. Only E. tapacumensis (30 μg/mL) caused damage DNA (p<0,05) in the alkaline version. At the neutral pH version, all concentrations tested (p<0,05) caused DNA damage. Thus, is concluded that E. cuspidifolia and E. tapacumensis essential oils are cytotoxic in tumor cell lines, and they have anticancer potencial in HCT 116 cell line acting as cytotoxic and cytostatic depending on the time and the tested concentration. Only E. tapacumensis is genotoxic, in non-tumor cell line, however more specific studies are needed to determine whether this genotoxicity is reversible or if the mechanism of cytotoxic action of essential oil is related to the cell DNA damage. / O câncer é uma das maiores causas de morte no mundo. A pesquisa de produtos naturais constitui uma estratégia para o desenvolvimento de agentes anticancerígenos. Objetivou-se nessa pesquisa identificar os componentes químicos, investigar o potencial anticâncer in vitro e a genotoxicidade de óleos essenciais do gênero Eugenia. Para isso, foram utilizados os ensaios de citotoxicidade do alamar blue e o potencial hemolítico em eritrócitos de camundongos, como triagem inicial para a seleção das amostras com potencial citotóxico, a partir da determinação da CI50 das amostras. Posteriormente a seleção das amostras, buscouse avaliar o efeito anticâncer dos óleos essenciais e a sua genotoxicidade. Foram testados nove óleos essenciais, em concentração única (50 μg/mL) dos quais apenas dois - Eugenia cuspidifolia (1) e Eugenia tapacumensis (3)- foram selecionados para os testes posteriores. A CI50, em 72 horas, obtida para essas duas amostras variou entre 4,69 a 24,35 μg/mL e 6,22 a 26,17 μg/mL, respectivamente, entre as linhagens tumorais. Para a linhagem não tumoral, em 72 horas, os valores de CI50 foram de 25,51μg/mL para E. cuspidifolia (1) e 36,12 μg/mL para E. tapacumensis (3), além de não causarem hemólise a eritrócitos de camundongos. As espécies E. cuspidifolia (1) e E. tapacumensis (3) foram mais ativas para as linhagens de colorretal (HCT 116) e ovário (ES-2), e escolheu-se para os testes posteriores utilizar a linhagem HCT 116, por ser um dos tipos de câncer mais frequentes em diagnósticos e também em mortes a nível mundial. A constituição química dos óleos essenciais de E. cuspidifolia (1) e E. tapacumensis (3) foi investigada através de cromatografia gasosa. Foram identificados um total de 26 constituintes para as duas amostras. Os componentes majoritários foram óxido de cariofileno, α-copaeno, hepóxido de humuleno II e cis-nerolidol. Os óleos nas concentrações de 7,5, 15 e 30 μg/mL, foram testados no ensaio clonogênico e reduziram significativamente o número de colônias (p<0,05) na concentração de 7,5 e 15 μg/mL, quando comparado ao controle DMSO (0,2%). No ensaio de motilidade celular, nos tempos de 24 e 48 horas, os óleos essencias reduziram a migração (p<0,05) somente na concentração de 30 μg/mL, no tempo de 48h. O potencial inibidor de metaloproteinases MMP-9 e MMP-2 foi avaliado através de método zimográfico. Nos resultados obtidos, o melhor efeito inibitório foi do óleo essencial de E. tapacumensis (3), o qual reduziu a atividade enzimática (p<0,05) nas concentrações de 15 e 30 μg/mL, nos dois tempos de tratamento (24 e 48h). A genotoxicidade foi avaliada pelo ensaio do cometa, usando duas versões, a de pH alcalino, o qual detecta qualquer dano causado ao DNA, e a de pH neutro, específico para quebras da fita dupla de DNA. Através da análise do Índice de dano, os resultados no ensaio em pH alcalino foram semelhantes ao do pH neutro. Somente E. tapacumensis (3) (30 μg/mL) causou dano ao DNA (p<0,05) na versão alcalina. Na versão em pH neutro todas as concentrações causaram dano (p<0,05). Assim, conclui-se que os óleos essenciais de E. cuspidifolia (1) e E. tapacumensis (3) são citotóxicos nas linhagens tumorais, tem potencial anticâncer na linhagem HCT116 agindo como citotóxico e citostático dependendo do tempo e da concentração testada e somente a E. tapacumensis (3) é genotóxica, em células não neoplásicas, , entretanto estudos mais específicos precisam determinar se essa genotoxicidade é reversível ou se o mecanismo de ação citotóxica do óleo está relacionada a lesão do DNA das células.

No description available.

Ett cirkulärt samhälle strävar efter att efterlikna naturens kretslopp där material som blir över från en verksamhet används i en annan, för att minska uttag av jungfruliga resurser. Från verksamheter kopplade till vägbyggen är det främst geologiska material som hanteras. Bergmaterial kan innehålla sulfidmineral som vid oxidation kan producera syra. För att säkerställa att bergmaterialet kan användas på ett säkert sätt används olika predikteringsmetoder för att identifiera risker med materialet, projektet har studerat några av dessa predikteringsmetoder och resultatet visar att de ger mycket osäkra bedömningar av det studerade materialet. I detta examensarbete har 10 bergmaterial studerats med syfte att undersöka och diskutera predikteringsmetoder och hur dessa kan tillämpas och tolkas för sulfidförande berg.  Bergmaterialet har en kemiskt analyserad svavelhalt mellan 0,3 och 1,6 vikts-%. Undersökningen inkluderar statiska och kinetiska tester. De statiska tester som använts är ABA och NAGpH, mineralogiska studier, både kvalitativt och i optiskt mikroskop och kvantitativt i SEM, XRF-mätningar, även ett enkelt laktest har utförts på materialet. Kinetiskt fuktkammartest (40 veckor) har inom ramen för projektet påbörjats på tre av bergmaterialen. Resultaten av predikteringstesterna visar på en stor variation mellan materialen utifrån använda riktvärden. Bedömningen av syrapotentialen och neutralisationspotentialen visar också stora variationer utifrån nyttjade riktlinjer. Det finns behov att vidare undersöka vad det är som gör att predikteringsmetoderna ger det resultat de ger och vidare hur detta kan tolkas. Statiska predikteringstester visar inte heller någon tidsaspekt som är viktigt att betrakta vid prediktering av material. Kinetiska test, som fuktkammartest visar hur utlakningen varierar med tiden.  Hur provtagningen av materialet utförs har också en stor påverkan på resultatet, det är därför viktigt att fortsatt utreda hur en representativ provtagning kan utföras, och hur olika provtagningsmetoder påverkar analysresultaten / In a circular society that strives to close the cycle of resources, usage of rest material from one operation area to another plays a natural role. In operations connected to roadworks geological material are mainly handled. Geological material can contain sulfide minerals which can through oxidation produce acid. To ensure that the material can be used in constructions different prediction methods are used to estimate risks with the material. In this master thesis prediction methods has been studied on 10 different materials, the results shows that the studied prediction methods gives a big variation in results. 10 rock samples has been studied with the aim to investigate and discuss prediction methods for sulfidic rock.  The rock samples had a sulfur content of 0,3 to 1,6 weight- %. The studied prediction methods included static test as Acid Base Accounting (ABA) and Net Acid Generation (NAGpH), mineralogical studies on the material. A kinetic test, humidity cell test (HCT) was started up in this project but not completed. Full rock chemistry and a simple leaching test was also performed on the material. The results from the prediction test on the material shows a large variation between the materials and when compared to different guideline values interpreted by different authorities. Determination of the neutralization and acid potential of the material also present a difference. There is a need to further investigate how the results from the studied prediction methods should be interpreted, and which tests are appropriate to use depending on where the material are planned to be used. How the sampling and sample preparation is preformed will also affect the result from prediction test. Therefore it´s important to investigate how representative sampling can be done to fulfill the purpose and how different sampling and preparation methods affects the results from analysis.

Trust is one of the attitudes that can affect the intentions and the behavior of a human using a system. Misusing a system can have safety- as well as economic consequences, this is why it is important that a user develops calibrated trust towards a system. In this report, the research question is: how much trust cross country skiers have towards a physical intelligent tutoring system (PITS)? Six biological males (age 24 to 50) roller skied on a mechanical treadmill and received feedback from the PITS on a TV-screen. Experience of using the PITS was evaluated with an instrument in a semi-structured interview. The instrument measured the participants overall perceived trust (OPT) for the system, and the participants were asked to further their thoughts about the statements. The data was transcribed, coded, and categorized in a thematic analysis. The result showed that a majority of the participants had low OPT for the PITS, and the thematic analysis showed that the minority with higher levels of OPT focused on the choice of an elite skier as the reference skier. One of the problems with the instrument was that it was developed for evaluation in long term usage, and not first time usage as in this study. The result of this report can be used for further development of the PITS and a reminder on why trust needs to be considered when creating user experiences.

Background: Iron deficiency, anemia, and iron excess have been associated with altered hearing functions in children and adults. Animal studies suggest that iron deficiency during pregnancy negatively affect fetal auditory development. The relationship between maternal iron status and auditory functions in healthy term newborns has not been clearly elucidated among humans. The goal of this pilot study was to determine the relationship between markers of iron status during pregnancy and brainstem auditory function in healthy neonates. Methods: Pregnant women who gave birth at the Montfort Hospital were recruited to take part in this study (n=6). Within two weeks after birth, their newborn’s hearing function was assessed by wave amplitude, latency and inter-latency from the Auditory Brainstem Response (ABR) test. Markers of iron status, namely hemoglobin (Hb) and mean corpuscular volume (MCV), were collected retrospectively for the first and second trimester from the women’s medical chart. Results: Overall, no significant relationship was observed between maternal Hb and MCV concentrations and newborns auditory function. Although two significant Spearman correlations were observed (MCV and inter-latency I-V; r=0.87; p=0.005 and Hb and amplitude V; r=0.89; p=0.04), these findings may be due to chance because of multiple testing and the small sample size. Conclusion: Although iron is a key nutrient involved in the brain and auditory system development, we were not able to demonstrate a relationship between iron status during pregnancy and newborn hearing function. Prospective or intervention studies with a larger sample size and with more specific iron markers (ex. ferritin) are required to confirm these findings.

Cette thèse est consacrée au développement de méthodes numériques pour la simulation de plaques et coques fissurées. Pour ce problème, les méthodes classiques sont basées sur la Méthode des Elements Finis (MEF). En raison de la présence d'une singularité en fond de fissure, la MEF souffre de plusieurs défauts. Son taux de convergence n'est pas optimal. De plus, en cas de propagation de la fissure, le domaine doit être remaillé. Une nouvelle méthode d'éléments finis, introduite en 1999 et baptisée XFEM, permet de s'affranchir de ces inconvénients. Dans cette méthode, la base éléments finis est enrichie par des fonctions de forme spécifiques qui représentent la séparation du matériau et la singularité de fond de fissure. Ainsi, domaine et fissure sont indépendants et le taux de convergence est optimal. Dans cette thèse, on développe deux formulations XFEM adaptées à un modèle de plaques minces. Ces méthodes ont pu être implémentées dans la bibliothèque d'éléments finis Getfem++, et testées sur des exemples où la solution exacte est connue. L'étude d'erreur montre que la méthode XFEM possède un taux de convergence optimal, alors que la MEF montre une convergence plus lente. L'autre contribution de cette thèse concerne le calcul de Facteurs d'Intensité de Contraintes (FIC) : ces grandeurs indiquent le risque de propagation de la fissure. Nous proposons deux méthodes de calcul originales, basées sur nos formulations XFEM. La première méthode utilise l'intégrale-J, et la deuxième fournit une estimation directe, sans post-traitement.

La classification est un des nombreux domaines de la Fouille de Données qui vise à extraire l'information à partir de grands volumes de données en utilisant différentes techniques computationnelles de l'apprentissage, des statistiques et de la reconnaissance des formes. Une des deux approches fondamentales de la classification non supervisé (ou clustering) est la classification hiérarchique. Son but est de produire un arbre dans lequel les nœuds représentent des classes des objets analysés. Un des inconvénients principaux de la méthode ascendante hiérarchique la plus connue et la plus utilisée, la Classification Ascendante Hiérarchique (CAH), est le fait qu'on ne peut pas mettre en évidence des classes d'objets ayant des caractéristiques communes. Cette propriété se trouve par exemple dans les classes qui se recouvrent et qui ont été introduites et étudies dans les extensions de la CAH.<br /><br />Cette thèse porte sur une extension récente de la Classification Ascendante Hiérarchique, appelée Classification Ascendante 2-3 Hiérarchique et proposé par P. Bertrand en 2002, avec en vue son application au domaine de la Fouille de Données.<br />Les trois contributions majeures de cette thèse résident dans l'étude théorique des 2-3 hiérarchies (appelées aussi paired hierarchies), dans le nouvel algorithme de 2-3 CAH avec son implémentation et dans la première étude applicative de cette méthode dans deux domaines de la Fouille de Données.<br /><br />Notre étude théorique inclut la découverte de quatre nouvelles propriétés théoriques des 2-3 hiérarchies et les définitions des liens d'agrégation entre les classes pour ce type de structure. Ceci nous a aussi permis de mettre en évidence un cas spécial de fusion des classes et d'introduire une étape intermédiaire dans la construction des 2-3 hiérarchies. L'étude exhaustive et systématique des cas possibles nous a permis de formuler les meilleurs choix concernant le lien d'agrégation et l'indexation de la structure, avec en vue l'amélioration de la qualité des 2-3 hiérarchies.<br /><br />Dans un deuxième temps, basé sur notre étude et contributions théoriques, nous proposons un nouvel algorithme général de Classification Ascendante 2-3 Hiérarchique. Ceci représente la concrétisation de notre travail précédent, aboutissant à un algorithme performant, qui explore plusieurs possibilités du modèle 2-3 hiérarchique. Une analyse théorique de la complexité de notre algorithme a montré que la complexité a été réduite de O(n3) dans l'algorithme initial de 2-3 CAH à O(n2 log n) pour notre algorithme. Les comparaisons des 2-3 hiérarchies avec les hiérarchies classiques obtenues sur différents ensembles de données (réels et simulés), ont validé l'analyse de complexité par les temps d'exécution. En plus, des résultats très satisfaisants ont été obtenus en analysant la "qualité" des 2-3 hiérarchies comparées aux hiérarchies classiques : jusqu'au 50% de classes en plus et un gain maximum de 84% en utilisant l'indice de Stress.<br />Nous avons ensuite proposé un modèle orienté-objet de notre algorithme de 2-3 CAH, qui a été intégré dans une boite à outils ``Hierarchical Clustering Toolbox'' (HCT) que nous avons développée pour la visualisation des méthodes ascendantes hiérarchiques de classification. Ce modèle a été également intégré comme méthode d'indexation des cas dans la plateforme de Raisonnement à Partir de Cas (RàPC), CBR*Tools, développé à l'INRIA Sophia Antipolis, et utilisé pour la conception des systèmes de recommandations. <br /><br />Notre dernière contribution concerne une toute première étude de l'utilisation de notre algorithme de 2-3 CAH sur des données réelles relevant de deux domaines de la Fouille des Données : le Web Mining et la Classification de Documents XML. Celle-ci a donné lieu à des résultats intéressants et portait sur la comparaison de la classification 2-3 hiérarchique des équipes de recherche de l'INRIA en utilisant soit le comportement des utilisateurs sur leur sites Web, soit leur rapport annuel d'activité écrit en XML, par rapport à la structure organisationnelle existante en thèmes de recherche.<br /><br />Pour conclure, nous montrons que ce sujet est loin d'être épuisé et nous proposons plusieurs pistes de recherche future relatives à la Classification Ascendante 2-3 Hiérarchique ainsi qu'à notre boite à outils HCT, développée pendant cette thèse.