Les stratégies actuelles de lutte contre le cancer consistent à développer de nouveaux traitements à base de molécules d’origine naturelle susceptibles de déclencher l’apoptose de cellules malignes et d’inhiber les principales voies de survie cellulaire. Le premier objectif de ces travaux porte sur les effets anti-prolifératifs et apoptotiques de l’extrait éthanolique de la plante Berberis libanotica sur les lignées érythroleucémiques humaines HEL, K562 et K562 (COX-2+) et sur son effet sur l’expression de la COX-2 dans ces lignées. Nos résultats montrent que l’extrait induit l’apoptose dans les lignées étudiées et ceci par activation de la caspase-3, le clivage de la PARP-1 et la fragmentation de l’ADN. De même, il induit la diminution de l’expression de COX-2. Nous avons démontré que les voies survie de NF-ĸB et p-AKT sont inhibées. Le deuxième objectif de ces travaux consistent à identifier la molécule présente dans cet extrait et qui est capable de déclencher ces effets anticancéreux. Nous avons démontré que la berbérine est la molécule majoritaire dans cet extrait et possède des effets apoptotiques et des effets inhibiteurs des voies de survie, ce qui est similaire aux effets de l’extrait brut.Mots clés :érythroleucémie, apoptose, berberis libanotica, berbérine, COX-2, NF-kB, p-AKT. / The first aim of this study focuses on the apoptotic effect of Berberis Libanotica on human erythroleukemia cell lines HEL, K562, and K562(COX-2+) and it is effect on the expression of COX-2. In light of the reported chemopreventive and chemosensitive effects of natural products on various tumor cells and animal models, we postulated that our Bl extract may mediate their effects through apoptosis induction with suppression of cell survival pathways. We showed that this extract induces apoptosis in eryrhtroleukemia cells by activation of the late markers of caspase-3 activation, PARP cleavage and DNA fragmentation. In the other hand, we showed that Bl extract reduced significantly expression of COX-2 by a dose-dependent manner. In regard to our results, it is clear that the simultaneous inhibition of Akt and NF-κB signalling can significantly contribute to the anticancer effects of Bl extract in human erythroleukemia cells.The second objective of this report is to elucidate wich molecule present in our extract can exert this effects. We found that berberine, the majoritory compound, can induce an apoptotic effect and inhibits the survival pathway of NF-ĸB and p-AKT similarly to the extract.Key words: erythroleukemia, apoptosis, Berberis libanotica ,berberine, COX-2, NF-ĸB, p-AKT

T. gondii interfère avec l'activation des voies de signalisation de NF-kB des cellules hôtes. Ainsi, lors de l’infection par T. gondii, 85% des gènes dépendant de NF-kB sont up-régulés. Un autre facteur de transcription dont l’expression est modulée par le parasite est UHRF1 (Ubiquitin-like,containing PHD and RING finger domains, 1). UHRF1, en se fixant sur le promoteur du gène de la cycline b, induit une répression épigénétique de ce dernier conduisant à un arrêt du cycle cellulaire des cellules infectées en phase G2 et à un arrêt de la prolifération parasitaire. L’analyse in silico du promoteur du gène uhrf1 a montré qu’il possédait 9 sites de fixation de NF-kB. Effectivement nous avons démontré que NF-kB interagit avec le promoteur du gène uhrf1 lors d’une infection par T. gondii. L’expression d’UHRF1 serait donc modulée par NF-kB dans les cellules infectées par T. gondii. Or NF-kB a une régulation différentielle en fonction de la nature de la souche infectante. Là encore, nous avons pu observer une régulation différentielle d’UHRF1 selon la nature de la souche infectante, pouvant être dues à la régulation souche dépendante de NF-kB. La détermination du rôle précis de l’activation d’UHRF1 dans les cellules infectées et l’identification du ou des facteurs parasitaires responsables pourraient permettre de mieux comprendre les mécanismes de persistance intracellulaire du parasite et de découvrir de nouveaux points d’impact thérapeutiques. / T.gondii interferes with the activation of NF-kB signaling pathways. Thus, upon infection by T.gondii, 85% of genes NF-kB-dependent are up-regulated. Another transcription factor whose expression is modulated by the parasite is UHRF1 (Ubiquitin-like, Containing PHD and RINGfinger domains, 1). UHRF1, bind to the gene promoter of cyclin b and induces epigenetic repression of this gene leading to cell cycle arrest in G2 phase of infected cells and stop the proliferation in both infected cells and parasite. In silico analysis of the uhrf1 gene promoter has been shown to possess 9 binding sites of NF-kB. Our study showed that NF-kB actually interacts with the promoter of gene uhrf1 during infection with T. gondii. This suggests that the expression of UHRF1 is modulated by NF-kB in T. gondii-infected cells. In addition we observed differential regulation of UHRF1 depending on the nature of the infecting strain. These variations may also be due to already well-known differential regulation of NF-kB by different strains of T.gondii. Determining the precise role of UHRF1 activation in infected cells and the identification of the parasitic factor responsible of this activation would allow to a better understanding of the mechanisms of intracellular persistence of the parasite and allow to unravel new therapeutic trails.

A musculatura esquelética tem um papel fundamental para a manutenção da homeostase do organismo. A perda de massa muscular está relacionada a prejuízos na qualidade de vida de indivíduos saudáveis, além de piorar o prognóstico de pacientes com doenças sistêmicas, como o câncer, o diabetes e a insuficiência cardíaca. Em quadros mais graves de insuficiência cardíaca, a perda excessiva de massa muscular associada a um reduzido consumo de oxigênio de pico, são considerados como preditores independentes de mortalidade. O aumento do estresse oxidativo tem sido apontado como um dos principais desencadeadores do aumento da degradação de proteínas na atrofia muscular. Na presente tese, investigamos a contribuição do estresse oxidativo para a ativação das vias de sinalização NF-kB, FOXO e MAPK na atrofia muscular desencadeada pela insuficiência cardíaca. Para compreender melhor os mecanismos envolvidos na ativação dessas vias pelo estresse oxidativo, utilizamos a linhagem de células musculares C2C12. Observamos que o tratamento com peróxido de hidrogênio (1,2mM, 12h) induziu um aumento do estresse oxidativo, o qual foi capaz de aumentar a atividade do proteassoma, desencadeando a atrofia dos miotúbulos. Verificamos também um aumento da expressão proteica de alguns componentes dessas vias de sinalização, como p-p38 e NF-kB; apontando para uma ativação diferenciada dessas vias pelo estresse oxidativo. Para verificar se essas vias de sinalização relacionadas ao estresse oxidativo estavam também relacionadas à atrofia desencadeada pela insuficiência cardíaca, avaliamos um modelo experimental de ratos com insuficiência cardíaca induzida pelo infartado do miocárdio. Observamos uma redução da área de secção transversa do músculo plantar, acompanhada de um aumento da inflamação sistêmica, de p38 e das atividades de NF-kB e do proteassoma. Como o treinamento físico aeróbico tem se apresentado como uma estratégia terapêutica não farmacológica eficaz na redução do estresse oxidativo e no restabelecimento da atividade do sistema ubiquitina proteassoma, submetemos os ratos infartados ao treinamento físico aeróbico em esteira rolante. O treinamento físico aeróbico preveniu a perda de massa muscular, reduzindo a inflamação sistêmica e as atividades de NF-kB e do proteassoma. Em conjunto, os resultados apontam para o estresse oxidativo como um fator preponderante para o aumento da degradação de proteínas relacionada à atrofia muscular, seja por indução de inflamação (TNF-α) ou por sua ação direta. Além disso, observamos que as vias de sinalização são ativadas de forma diferenciada nos dois modelos, sugerindo que a degradação de proteínas nos miotúbulos está relacionada ao controle de qualidade de proteínas e, nos ratos infartados, às alterações do metabolismo, servindo como fonte de energia. Já o treinamento físico aeróbico comprovou sua eficácia no restabelecimento da atividade do proteassoma, reduzindo a inflamação e a atividade de NF-kB, prevenindo assim, a perda de massa muscular / About 40% of human body mass consists of skeletal muscles, which are involved in all aspects of movement including breathing, eating, posture, walking and reflexes. Skeletal muscle is also important as a source of heat generation and as a regulator of intermediary metabolism. Loss of skeletal muscle mass and function (skeletal muscle atrophy) leads to several functional impairments, affecting health and quality of life. It occurs in several chronic diseases such as cancer, diabetes and heart failure. In heart failure, atrophy is considered an independent predictor of poor prognosis. Oxidative stress has a crucial role in atrophy, activating different signaling pathways capable of stimulating the ubiquitin proteasome system to degrade proteins. In this study, we investigated the oxidative stress contribution to NF-kB, FOXO and MAPK signaling pathway activation in heart failure-induced atrophy. To better understand the mechanisms involved with oxidative stress and signaling pathways activation in atrophy, we have used C2C12 skeletal muscle cells. We observed that, even in high hydrogen peroxide concentrations, oxidative stress increased proteasome activity, phosphorylated p38 and NF-kB protein expression, causing myotubes atrophy. In an experimental heart failure model of infarcted rats, we evaluated plantaris muscle and verified a reduced cross sectional area, accompanied by increased systemic inflammation, p-38 protein expression and increased both NF-kB and proteasome activities. As aerobic exercise training causes a lot of beneficial effects on skeletal muscle structure and function in chronic diseases, we submitted infarcted rats to 8 weeks of aerobic exercise training on a treadmill. Aerobic exercise training prevented atrophy by reducing inflammation and both NF-kB and proteasome activities. Collectively, our data suggest a differentiated activation by oxidative stress in muscle cells and animal models. In the first case, protein degradation was involved with protein quality control; and, in the other, oxidative stress is a second messenger, stimulating protein degradation to provide substrates to metabolism. Aerobic exercise training re-established proteasome activity by reducing inflammation and NF-kB activity, preventing muscle atrophy

La voie de signalisation NF-kB (nuclear factor kB) régule la transcription de gènes impliqués dans la réponse immune et l’inflammation. L’activation chronique de cette voie est fréquemment retrouvée associée à des désordres inflammatoires et des cancers. Les impacts fonctionnels de l’activation de la voie NF-kB ont été jusqu’à présent étudiés à l’échelle des promoteurs. Néanmoins, les études récentes de la distribution chromatinienne de NF-kB indiquent que la sous-unité NF-kB RelA se localise majoritairement dans les régions intragéniques, incluant des exons et des introns, où ses fonctions restent inconnues. Mes travaux ont consisté à adresser cette question dans le contexte de l’infection par le virus HTLV-1, un activateur chronique de la voie NF-kB, responsable de la leucémie T de l’adulte. Mes données montrent que l’activation de la voie NF-kB par l’oncogène viral Tax de HTLV-1 s’accompagne de modifications de l’épissage alternatif d’exons riches en GC qui coïncident avec le recrutement chromatinien de RelA à proximité de ces exons régulés. Les analyses intégratives des profils d’épissage et du remodelage de la chromatine, combinées à des essais de ciblage expérimental de la chromatine (TALE), démontrent que la fixation intragénique de RelA permet de recruter le régulateur d'épissage DDX17 pour moduler l’épissage alternatif de l’exon via son activité hélicase. Ces données révèlent que, outre ses fonctions transcriptionnelles, le facteur NF-kB RelA agit comme une ancre chromatinienne pour le facteur d’épissage DDX17 et fournit une spécificité de régulation d’épissage alternatif. Ces données revisitent nos connaissances des mécanismes physiopathologiques des maladies associées à HTLV-1 ainsi que d'autres désordres reliés à l’activation chronique de la voie NF-kB / The NF-kB (nuclear factor kB) signaling pathway regulates gene transcription of genes involved in immune response and inflammation. Chronic activation of NF-kB frequently associated with inflammatory disorders and cancer. The functional impacts of NF-kB have long been studied at the promoter level. Nevertheless, recent studies of the chromatin distribution of RelA indicate that this NF-kB subunit is predominantly localized in intragenic regions, including exons and introns, where its functions remain unknown. My work has addressed this question in the context of HTLV-1 infection, which is a constitutive activator of NF-kB, and the causative agent of the Adult T-cell Leukemia. The results show that the activation of NF-kB by the viral oncoprotein Tax results in changes in alternative splicing regulations of GC-rich exons that coincide with the chromatin recruitment of RelA in the vicinity of these exons. Integrative analysis of RNA splicing and chromatin occupancy, combined with experimental chromatin tethering assays (TALE) demonstrate that the intragenic binding of RelA leads to the recruitment of the splicing regulator DDX17, which modulates the inclusion rate of exon thanks to its helicase activity. Altogether, these data reveal that, besides its transcriptional role, NF-kB RelA acts as a chromatin anchor for the splicing factor DDX17 and provides alternative splicing specificity. These data revisit our knowledge of the physiopathologic mechanisms of HTLV-1 associated diseases , as well as other disorders related to the chronic activation of the NF-kB pathway

Die Gliazellen des enterischen Nervensystems galten lange Zeit lediglich als Packmaterial der Neuronen. Erst in den letzten Jahren rückte dieser Zelltyp in das Interesse der Forschung, weil mehrere Studien eine Schlüsselrolle bei der Erhaltung der Darmwandintegrität postulierten. Da bisher wenig über die immunphysiologischen Eigenschaften der enterischen Gliazellen bekannt war, wurde in dieser Arbeit deren Bedeutung bei der angeborenen intestinalen Immunantwort untersucht. Hierfür wurden Primärkulturen von enterischen Gliazellen eingesetzt, welche aus dem Plexus myentericus adulter Ratten stammten. Es wurde erstmalig nachgewiesen, dass enterische Gliazellen über Toll like und NOD-Rezeptoren (TLR 2, TLR 4, TLR 6, TLR 7, TLR 9, NOD 1, NOD 2) pathogene bakterielle Muster (PAMPs) erkennen. Dabei zeigte sich, dass eine enge Vernetzung zwischen den Toll-like-Rezeptoren besteht. Einerseits findet eine Liganden-spezifische Regulierung der Toll-like-Rezeptoren statt, anderseits beeinflussen TLR spezifische Liganden die mRNA-Expression weiterer Toll-like-Rezeptoren. In der Analyse der intrazellulären Signalweiterleitung konnte die Existenz verschiedener Adaptor- und Interaktionsmoleküle wie RICK und Myd88 sowie der IKK / NF κB Signalweg mit Degradierung von IκB α nach Aktivierung durch LPS belegt werden. Aktivierte enterische Gliazellen sind eine Quelle für pro- und anti-inflammatorische Interleukine (IL-1α, IL 1β, IL 6, IL-10, IL-12), TNF α und Chemokine (Ccl-2, Cxcl-9, Cxcl-10). Sie können somit weitere immunkompetente Zellen rekrutieren und agieren vermutlich als Schnittstelle zwischen der angeborenen und erworbenen Immunantwort. Die Ergebnisse dieser Arbeit implizieren, dass es sich bei enterischen Gliazellen um einen zusätzlichen immunregulatorischen Zelltyp im Darm handelt, welcher aktiv bei der angeborenen Immunabwehr mitwirkt. Neben pro inflammatorischen Eigenschaften tragen enterische Gliazellen auch zur Darmprotektion während Entzündungsprozessen bei. Daraus ergibt sich eine komplexe Funktionalität der enterischen Gliazellen bei der immunologischen Homöostase im Darm.

Es wird vermutet, dass das Zusammenspiel von NF-κB-Signalen und Autophagie die Entzündung in verschiedenen zellulären Kontexten und als Reaktion auf unterschiedliche Stimuli reguliert. Der molekulare Mechanismus, durch den diese beiden Signalwege bei der Regulierung der Entzündungsreaktion zusammenwirken, ist jedoch noch nicht bekannt. Mithilfe biochemischer Analysen und bildgebender Verfahren haben wir zum ersten Mal die Interaktion zwischen dem autophagischen Marker LC3 und der NF-κB/p65-Untereinheit als Reaktion auf verschiedene Stressbedingungen charakterisiert. Wir konnten zeigen, dass die Anhäufung von LC3 im Zellkern nach der NF-κB-Aktivierung mit p65 interagiert, was durch die Ubiquitinierung des p65-Proteins gefördert und durch den Cargo-Rezeptor p62 erkannt wird. Zusammengenommen weisen diese Daten auf eine neue Rolle von p62 beim Transport von im Kern ubiquitiniertem p65 zu Autophagosomen hin, wo es abgebaut wird, um die entzündungsbedingte NF-κB-Hyperaktivierung zu kontrollieren. Diese Erkenntnisse sind wichtig für die Entwicklung neuer therapeutischer Strategien gegen Krankheiten, die mit einer gestörten Autophagie und einer konstitutiven NF-κB-Aktivität einhergehen. Die NF-κB-Signalübertragung spielt nicht nur eine entscheidende Rolle bei Entzündungen und der Tumorbildung, sondern ist auch für Entwicklungsprozesse wichtig. Durch die Etablierung von 3D-Organoid-Kulturen aus dem Dünndarm und unter Verwendung verschiedener Mauslinien weisen wir im zweiten Teil der Arbeit nach, dass NF-κB eine wichtige Funktion bei der Zelldifferenzierung und der Erhaltung von Stammzellen in vivo und ex-vivo spielt. Wir konnten zeigen, dass die Proliferation und das Absterben von Darmepithelzellen (IEC) bei Mäusen mit ubiquitärer Unterdrückung der NF-κB-Aktivität unverändert sind, während die Zahl der Becherzellen auf Kosten der Paneth-Zellen zunimmt. Zusammenfassend lässt sich sagen, dass unsere Ergebnisse eine neue IEC-immanente Rolle von NF-κB bei Entscheidungen über das Zellschicksal und die Differenzierung aufzeigen, die über die Regulierung der Wnt-Signale und der Sox9-Expression stromabwärts von NF-κB erfolgt. Die hier beschriebenen Erkenntnisse verbessern unser Verständnis der NF-κB-Funktionen in der Stammzellbiologie, die, wenn sie dereguliert sind, auch Auswirkungen auf die Entzündung des Darms und die Tumorentstehung haben. / The interplay between NF-κB signaling and autophagy has been suggested to regulate inflammation in different cellular contexts and in response to different stimuli. However, the molecular mechanism by which these two pathways interact to regulate the inflammatory response remains elusive. By using biochemical analysis and imaging techniques, we characterized for the first time the interaction of autophagic marker LC3 and NF-κB/p65 subunit in response to different stress conditions. We demonstrated that the accumulation of LC3 within the nucleus interacts with p65 following NF-κB activation, which is promoted by ubiquitination of p65 protein and recognized by the cargo receptor p62. Together, these data identify a novel role for p62 in trafficking nuclear-ubiquitinated p65 to autophagosomes for degradation to control inflammation-driven NF-κB hyperactivation. These findings are important for developing novel therapeutic strategies against diseases involving defective autophagy and constitutive NF-κB activity. In addition to its critical role in inflammation and tumor formation, NF-κB signaling is essential in developmental processes. Establishing 3D organoid culture from the small intestine and using different mouse lines, we prove in the second part of the thesis that NF-kB plays an important function in cell differentiation and stem cell maintenance in vivo and in ex-vivo. We demonstrated that while intestinal epithelial cell (IEC) proliferation and death are unaltered in mice with ubiquitous suppression of NF-κB activity, goblet cell numbers increase at the expense of Paneth cells. In summary, our results revealed a novel IEC-intrinsic role of NF-κB in cell fate decisions and differentiation which occur via regulation of Wnt signaling and Sox9 expression downstream of NF-κB. The findings described here improve our understanding of NF-κB functions in stem cell biology which, when deregulated, also have an impact on intestinal inflammation and tumorigenesis.

This work explores the capabilities of KB-BERT on the downstream task of Question Classification. The TREC data set for Question Classification with the Li and Roth taxonomy was translated to Swedish, by manually correcting the output of Google’s Neural Machine Translation. 500 new data points were added. The fine-tuned model was compared with a similarly trained model based on Multilingual BERT, a human evaluation, and a simple rule-based baseline. Out of the four methods of this work, the Swedish BERT model (SwEAT- BERT) performed the best, achieving 91.2% accuracy on TREC-50 and 96.2% accuracy on TREC-6. The performance of the human evaluation was worse than both BERT models, but doubt is cast on how fair this comparison is. SwEAT-BERTs results are competitive even when compared to similar models based on English BERT. This furthers the notion that the only roadblock in training language models for smaller languages is the amount of readily available training data. / Detta arbete utforskar hur bra den svenska BERT-modellen, KB-BERT, är på frågeklassificering. BERT är en transformermodell som skapar kontextuella, bidirektionella ordinbäddningar. Det engelska datasetet för frågeklassificering, TREC, översattes till svenska och utökades med 500 nya datapunkter. Två BERT-modeller finjusterades på detta nya TREC-dataset, en baserad på KB-BERT och en baserad på Multilingual BERT, en flerspråkig variant av BERT tränad på data från 104 språk (däribland svenska). En regel-baserad modell byggdes som en nedre gräns på problemet, och en mänsklig klassificeringsstudie utfördes som jämförelse. BERT-modellen baserad på KB-BERT (SwEAT-BERT) uppnådde 96.2% korrekthet på TREC med 6 kategorier, och 91.2% korrekthet på TREC med 50 kategorier. Den mänskliga klassificeringen uppnådde sämre resultat än båda BERT-modellerna, men det är tvivelaktigt hur rättvis denna jämförelse är. SwEAT-BERT presterade bäst av metoderna som testades i denna studie, och konkurrenskraftigt i jämförelse med engelska BERT-modeller finjusterade på det engelska TREC-datasetet. Detta resultat stärker uppfattningen att tillgänglighet till träningsdata är det enda som står i vägen för starkare språkmodeller för mindre språk.

This master thesis aims to determine if a Swedish BERT model can improve a BM25 search by re-ranking the top search results. We compared a standard BM25 search algorithm with a more complex algorithm composed of a BM25 search followed by re-ranking the top 10 results by a BERT model. The BERT model used is KB-BERT, a publicly available neural network model built by the National Library of Sweden. We fine-tuned this model to solve the specific task of evaluating the relevancy of search results. A new Swedish search evaluation dataset was automatically generated from Wikipedia text to compare the algorithms. The search evaluation dataset is a standalone product and can be beneficial for evaluating other search algorithms on Swedish text in the future. The comparison of the two algorithms resulted in a slightly better ranking for the BERT re-ranking algorithm. These results align with similar studies using an English BERT and an English search evaluation dataset. / Denna masteruppsats syftar till att avgöra om en svensk BERT-modell kan förbättra en BM25-sökning genom att ranka om de bästa sökresultaten. Vi jämförde en standard BM25-sökalgoritm med en mer komplex algoritm som består av en BM25-sökning följt av omrankning av de 10 bästa resultaten med en BERT-modell. BERT-modellen som används är KB-BERT, en allmänt tillgänglig neural nätverksmodell byggd av Kungliga biblioteket. Vi finjusterade den här modellen för att lösa den specifika uppgiften att utvärdera sökresultatens relevans. En ny svensk datamängd för utvärdering av sökresultat genererades automatiskt från Wikipedia-text för att jämföra algoritmerna. Datamängden är en fristående produkt och kan vara till nytta för att utvärdera andra sökalgoritmer på svensk text i framtiden. Jämförelsen av de två algoritmerna resulterade i en något bättre rankning för BERT-omrankningsalgoritmen. Dessa resultat överensstämmer med liknande studier som använder en engelsk BERT och en engelsk datamängd för utvärdering av sökresultat.

Presented here is a method for automatic punctuation restoration in Swedish using a BERT model. The method is based on KB-BERT, a publicly available, neural network language model pre-trained on a Swedish corpus by National Library of Sweden. This model has then been fine-tuned for this specific task using a corpus of government texts. With a lower-case and unpunctuated Swedish text as input, the model is supposed to return a grammatically correct punctuated copy of the text as output. A successful solution to this problem brings benefits for an array of NLP domains, such as speech-to-text and automated text. Only the punctuation marks period, comma and question marks were considered for the project, due to a lack of data for more rare marks such as semicolon. Additionally, some marks are somewhat interchangeable with the more common, such as exclamation points and periods. Thus, the data set had all exclamation points replaced with periods. The fine-tuned Swedish BERT model, dubbed prestoBERT, achieved an overall F1-score of 78.9. The proposed model scored similarly to international counterparts, with Hungarian and Chinese models obtaining F1-scores of 82.2 and 75.6 respectively. As further comparison, a human evaluation case study was carried out. The human test group achieved an overall F1-score of 81.7, but scored substantially worse than prestoBERT on both period and comma. Inspecting output sentences from the model and humans show satisfactory results, despite the difference in F1-score. The disconnect seems to stem from an unnecessary focus on replicating the exact same punctuation used in the test set, rather than providing any of the number of correct interpretations. If the loss function could be rewritten to reward all grammatically correct outputs, rather than only the one original example, the performance could improve significantly for both prestoBERT and the human group. / Här presenteras en metod för automatisk återinföring av skiljetecken på svenska med hjälp av ett neuralt nätverk i formen av en BERT-modell. Metoden bygger på KB-BERT, en allmänt tillgänglig språkmodell, tränad på ett svensk korpus, av Kungliga Biblioteket. Denna modell har sedan finjusterats för den här specifika uppgiften med hjälp av ett korpus av offentliga texter från landsting och dylikt. Med svensk text utan versaler och skiljetecken som inmatning, ska modellen returnera en kopia av texten där korrekta skiljetecken har placerats ut på rätta platser. En framgångsrik modell ger fördelar för en rad domäner inom neurolingvistisk programmering, såsom tal- till- texttranskription och automatiserad textgenerering. Endast skiljetecknen punkt, kommatecken och frågetecken tas i beaktande i projektet på grund av en brist på data för de mer sällsynta skiljetecknen såsom semikolon. Dessutom är vissa skiljetecken någorlunda utbytbara mot de vanligaste tre, såsom utropstecken mot punkt. Således har datasetets alla utropstecken ersatts med punkter. Den finjusterade svenska BERT-modellen, kallad prestoBERT, fick en övergripande F1-poäng på 78,9. De internationella motsvarande modellerna för ungerska och kinesiska fick en övergripande F1-poäng på 82,2 respektive 75,6. Det tyder på att prestoBERT är på en liknande nivå som toppmoderna motsvarigheter. Som ytterligare jämförelse genomfördes en fallstudie med mänsklig utvärdering. Testgruppen uppnådde en övergripande F1-poäng på 81,7, men presterade betydligt sämre än prestoBERT på både punkt och kommatecken. Inspektion av utdata från modellen och människorna visar tillfredsställande resultat från båda, trots skillnaden i F1-poäng. Skillnaden verkar härstamma från ett onödigt fokus på att replikera exakt samma skiljetecken som används i indatan, snarare än att återge någon av de många korrekta tolkningar som ofta finns. Om loss-funktionen kunde skrivas om för att belöna all grammatiskt korrekt utdata, snarare än bara originalexemplet, skulle prestandan kunna förbättras avsevärt för både prestoBERT såväl som den mänskliga gruppen.

Doctor of Philosophy / Department of Biochemistry / Lawrence C. Davis / Halogenated organic compounds have had widespread and massive applications in industry, agriculture, and private households, for example, as degreasing solvents, flame retardants and in polymer production. They are released to the environment through both anthropogenic and natural sources. The most common chlorinated solvents present as contaminants include tetrachloroethylene (PCE, perchloroethene), trichloroethene (TCE), trichloroethane (TCA), and carbon tetrachloride (CT). These chlorinated solvents are problematic because of their health hazards and persistence in the environment, threatening human and environmental health. This contribution provides insight on PCE degradation at laboratory and field scale at a former dry cleaning site in Manhattan, KS. Biostimulation experiments included combinations and concentrations of the following nutrients: soy oil methyl esters (SOME), yeast extract (YE), glucose, lactate, methanol and cheese whey. Bioaugmentation studies used KB-1 bacterial consortium (commercially available culture containing Dehalococcoides). This culture is known to complete the degradation of PCE to a safe end product, ethene. Concentrations of PCE and its degradation intermediates were monitored in the gas phase of the microcosm vials. Biostimulation of the natural ground water and soil microflora did not completely degrade PCE as cis-DCE (c-DCE) accumulated in the sample. Bioaugmented microcosms containing YE and SOME created reducing conditions for KB-1 culture, resulting in ~ 90% dechlorination of PCE to methane and c-DCE. Cheese whey microcosms containing 0.05% cheese whey inhibited the KB-1 culture. This inhibition was due to a drop of pH that inhibited the culture activity. Lower concentrations of cheese whey (e.g. 0.01% to 0.025%) reduced PCE and generated methane in KB-1 augmented microcosms. Based on microcosm results, a pilot bioremediation field study was conducted for a dry cleaning site contaminated with PCE. Ground water flow threatened public water wells located 1.5 miles from the source. Concentrations of PCE in the aquifer was 15 mg/L above the maximum contaminant level of 5 µg/L. Tracer studies with potassium bromide (KBr) were conducted before, during and after the bioremediation study. Nutrient solutions prepared with YE, SOME, lactate and glucose were used for biostimulation and preconditioning of ground water prior to KB-1 injection. Nutrients were provided twice during the pilot study to supplement microbial growth and cheese whey was used. During biostimulation no degradation beyond DCE was evident. The addition of KB-1 reduced PCE and DCE concentrations in the monitoring wells of the pilot study area. Total chlorinated ethene concentrations did not reach background levels 2 years after the last nutrient addition. Tracer results showed that microbial growth decreased ground water velocity during the study, but returned to normal conditions 1 year after the last nutrient addition. In this study we were able to show that native microbial population was not able to degrade PCE to final end products. Therefore, it was necessary to introduce KB-1 culture a long with nutrients to support complete reductive dechlorination of PCE.