The prefrontal cortex (PFC) is critically involved in many higher cognitive functions such as goaldirected behaviour, affective behaviour and especially working memory. In vivo extracellular recordings of PFC neural activity during working memory tasks show high variety in observed spiking patterns. These complex dynamics are critically shaped by intrinsic, synaptic and structural parameters of respective prefrontal networks. Moreover, dopamine (DA) is crucial for correct functioning of the PFC during working memory tasks. DA modulates a number of synaptic and intrinsic biophysical properties of single neurons, in particular deep layer pyramidal cells, which represent the major output neurons of the PFC. Despite a high variability of cortical pyramidal cell firing patterns, and somatodendritic morphology, no study has yet systematically examined correlations between intrinsic properties, morphological features and dopaminergic modulation of intrinsic properties. This study investigated properties of deep layer pyramidal cells through whole cell patch clamp in acute brain slices of the adult rat PFC. Cells were characterised physiologically through a variety of stimulation protocols surveying different time scales and wide intensity ranges, while all fast synaptic transmission was blocked. Furthermore the same catalogue of stimuli was recorded whilst applying specific DA receptor agonists to elucidate effects of DA receptor activation on intrinsic properties. All recorded cells were injected with biocytin and dendritic morphology was reconstructed from confocal image stacks of fluorescently labelled neurons. From the resulting data a set of characteristic variables were defined and a combination of principal components analysis and hierarchical cluster analysis was used to identify similarity between recorded cells in different parameter spaces spanned by intrinsic properties, intrinsic properties under dopaminergic modulation and morphology, respectively. The analysis presents evidence for distinct subpopulations within prefrontal deep layer pyramidal cells, as seen by clustering of recorded cells in these high dimensional parameter spaces. These subpopulations also show distinct input-output relationships, bearing implications for computational functions of these subpopulations. Furthermore, this study presents for the first time evidence of subpopulation specific DA effects in deep layer pyramidal cells. The quantitative analysis of somatodendritic morphology confirms physiological subpopulations and identifies characteristic morphological features of deep layer pyramidal cells. Moreover, cluster observed in different parameter spaces overlap, leading to a definition of subpopulations that concurs with previously described prefrontal pyramidal cell types. In conclusion, the results presented provide some deeper insight into fundamental principles of information processing in prefrontal pyramidal cells under the influence of dopamine.

La salinité du sol affecte la croissance des plantes glycophytes telle que Arabidopsis thaliana et le riz. Chez les plantes vasculaires, les racines sont composées de divers types de cellules organisées en cercles concentriques. Chaque type de cellules racinaires possède une fonction biologique spécifique et coordonnée avec les autres cellules qui composent cette même racine. Il est probable que la réponse des gènes au stress salin soit spécifique du type cellulaire, ce qui ne peut être révélé par des études à l'échelle de l'organe entier. Afin d'étudier les réponses spécifiques, notre approche a été de générer des profils de transcriptome pour deux types de cellules racinaires chez les plantes modèles, Arabidopsis et riz. Les deux types de cellules étudiées ont été choisis en raison de leur rôle possible soit dans le stockage du sodium dans les cellules corticales, soit dans son transport dans les cellules du péricycle chez Arabidopsis ou du cylindre central chez le riz. Des plantes exprimant la protéine fluorescente verte (GF) spécifiquement dans un type de cellule racinaire furent utilisées pour cette analyse. Les cellules ont donc pu être isolées chez le riz et Arabidopsis grâce à la technique de cytométrie en flux.L'analyse du transcriptome des cellules du péricycle et du cylindre central montrent que les cellules corticales sont plus réactives au stress salin et qu'une large majorité des gènes est sous–exprimée chez les deux plantes modèles. D'après les analyses d'expression des cellules du cortex d'Arabidopsis, trois voies métaboliques sont significativement sous-exprimées en réponse au stress salin: la voie de biosynthèse des phénylpropanoïdes, le transport de l'eau and le métabolisme secondaire. La régulation de gènes impliqués dans le transport de l'eau et des nutriments démontre l'importance des cellules corticales dans le mouvement des solutés. Chez le riz, les profils des deux types cellulaires étudiés révèlent une forte réaction de défense ; en effet le métabolisme protéique et la régulation de la transcription sont fortement sous-exprimés dans les cellules corticales alors que les cellules du cylindre central modifient et activent les gènes correspondant à divers catégories fonctionnelles telles que la réplication de l'ADN et le transport. Des gènes candidats ont été sélectionnés dans les deux types cellulaires d'Arabidopsis. Des lignées mutantes n'exprimant pas ces gènes ont été testées en stress salin dans des conditions hydroponiques. Les résultats ont révélé un phénotype accumulant moins de sodium dans les parties aériennes (20% par rapport au génotype sauvage) pour certaines de ces lignées mutantes. Ce travail est la première étude de transcriptome utilisant des types spécifiques de cellules racinaires chez le riz. L'identification de gènes et voies métaboliques répondant au stress salin dans le cortex et le cylindre central de la racine ouvre de nouveaux axes de recherche et va permettre d'élucider la complexité des processus biologiques d'adaptation au stress salin. / Soil salinity reduces the growth of glycophytic plants such as Arabidopsis thaliana and rice. In vascular plants, roots are organized into concentric layers of cells and each layer has a specific biological function coordinated with other cell types in the root. Therefore, genes differentially expressed in response to a salt stress are also likely to be changing only in specific cell types, and thus may not be revealed at the organ level. In order to identify novel salt-responsive genes, cell-type specific transcriptomic approaches were undertaken in Arabidopsis thaliana and rice, with application of physiologically reasonable salt stress (50mM) over 48 hours. Two cell-types from the root were chosen in both species for their potential role in salt storage and transport: cortical and pericycle/stelar cells respectively. Cell-types of interest expressing specifically Green Fluorescent Protein (GFP) were isolated from the rest of the root using fluorescence-activated cell sorting (FACS).The outer layer of the root was found to be responding more than the inner part of the root after 48 hours of salt stress, with an overall down-regulation in both rice and Arabidopsis. Arabidopsis cortical cells responding to salt seem to regulate the cell wall biosynthesis, which may modulate the shape of the cells or alter the apoplastic movements of solutes in response to salt. Genes related to transport were affected by salt in Arabidopsis, with the crucial role of cortical cells in the movement of solutes being evident. Rice cortical cells respond to salt by showing a more extreme defence reaction in changing the protein metabolism and the regulation of transcription. The response of the inner part of the rice root to 48 hours of mild salt stress showed up-regulation of genes implicated in broader functional categories. The biological relevance of genes revealed using cell-type specific transcriptomics was demonstrated in a salt assay using knock-out (KO) lines of candidate genes from both cell-types in Arabidopsis thaliana. Three KO mutant lines showed 20% reduction in shoot sodium after 5 weeks of salt stress and were also able to maintain a higher shoot dry weight. These transcriptomic studies of isolated stelar and cortical cells in response to a mild salt stress have revealed salt responsive genes and pathways, indicating new areas for further study, and contributing to our understanding of the complex responses of plants to their environment at the cellular level.

Single-cell RNA-sequencing (scRNA-seq) technology enables researchers to investigate a genome at the cellular level with unprecedented resolution. An organism consists of a heterogeneous collection of cell types, each of which plays a distinct role in various biological processes. Hence, the first step of scRNA-seq data analysis often is to distinguish cell types so that they can be investigated separately. Researchers have recently developed several automated cell type annotation tools based on supervised machine learning algorithms, requiring neither biological knowledge nor subjective human decisions. Dropout is a crucial characteristic of scRNA-seq data which is widely utilized in differential expression analysis but not by existing cell annotation methods. We present scAnnotate, a cell annotation tool that fully utilizes dropout information. We model every gene’s marginal distribution using a mixture model, which describes both the dropout proportion and the distribution of the non-dropout expression levels. Then, using an ensemble machine learning approach, we combine the mixture models of all genes into a single model for cell-type annotation. This combining approach can avoid estimating numerous parameters in the high-dimensional joint distribution of all genes. Using fourteen real scRNA-seq datasets, we demonstrate that scAnnotate is competitive against nine existing annotation methods, and that it accurately annotates cells when training and test data are (1) similar, (2) cross-platform, and (3) cross-species. Of the cells that are incorrectly annotated by scAnnotate, we find that a majority are different from those of other methods. / Graduate / 2023-07-27

No description available.

Une des questions fondamentales en neurobiologie du développement est de comprendre quels sont les mécanismes responsables de la formation des circuits neuronaux. Alors que les circuits de neurones projecteurs ont fait l'objet d'études intensives, aujourd'hui encore peu de chose est connu sur les molécules et les voies de signalisation impliquées dans la croissance et la différentiation axonale des interneurones GABA. Au cours de ma thèse, j'ai étudié le rôle de la molécule de guidage axonal Semaphorine 3a (Sema3A) dans le développement des axones des interneurones GABA du cervelet. Pour mener à bien cette étude, j'ai développé de nouvelles approches expérimentales in vitro et in vivo. En utilisant ces nouveaux outils, j'ai montré que la SEMA3A était capable d'attirer l'axone des interneurones GABA. In vitro, la SEMA3A induit une augmentation locale de la formation de branches axonales et l'apparition précoce d'un marqueur présynaptique (GAD65). Nous avons également montré que la combinaison d'expression de SEMA3A avec une molécule d'adhésion de la famille L1CAM dans les cellules hétérologues était capable d'induire l'innervation de cette cellule de manière spécifique par les interneurones GABA du cervelet. Cette étude nous a permis de montrer que la combinaison entre une molécule de guidage axonal et une molécule d'adhésion cellulaire spécifique était suffisante pour induire des mécanismes de reconnaissance cellulaire par les interneurones GABA du cervelet. / GABAergic interneurons are fundamental component in neural processing and their specific innervation patterns are though to be the building block for physiological brain function and computing. However the molecular and cellular mechanisms that assemble inhibitory local circuits remain largely unknown. In cerebellar cortex, molecular layer GABAergic interneurons are key regulators of cerebellar signal coding and memory formation by sending specifically their axons to innervate the Purkinje cells. Here, we show that a combination of both secreted axon guidance and recognition molecules of L1CAM family is sufficient to trigger target cell recognition by molecular layer GABAergic interneuron˙s in vivo. Using BAC transgenic reporter mice for cell-type specific gene-expression profiling of secreted SEMAPHORIN molecules, we identified that SEMAPHORIN3A (SEMA3A) expression picked precisely at relevant time-point of GABAergic local circuit formati on. In vitro, in a co-culture model, we found that semaphorin3A (SEMA3A) secreted by CHO cells attracts GABAergic interneurons axons and triggers their local specific branching. In vivo, the injection of these heterologous cells expressing SEMA3A ectopically in the granule cell layer is able to disrupt the « crystal »like organization of molecular GABAergic interneurons and attracts their axons in this ectopic territory. Moreover we found that both in vitro and in vivo, the co-expression of SEMA3A and the L1CAM family recognition molécules, Neurofascin, but not their respective expression alone, are able to induced heterologous cells innervation by molecular GABAergic interneurons. These results suggest that specific combination between axon guidance molecules and L1CAM family is sufficient to specify cell type recognition in a space and timely dependent manner.

With increasing amounts of data being collected in various fields of neuroscience, there is a growing need for robust techniques for the analysis of this information. This thesis focuses on the evaluation and development of quantitative frameworks for the analysis and classification of neuronal data from a variety of contexts. Firstly, I investigate methods for analysing spontaneous neuronal network activity recorded on microelectrode arrays (MEAs). I perform an unbiased evaluation of the existing techniques for detecting ‘bursts’ of neuronal activity in these types of recordings, and provide recommendations for the robust analysis of bursting activity in a range of contexts using both existing and adapted burst detection methods. These techniques are then used to analyse bursting activity in novel recordings of human induced pluripotent stem cell-derived neuronal networks. Results from this review of burst analysis methods are then used to inform the development of a framework for characterising the activity of neuronal networks recorded on MEAs, using properties of bursting as well as other common features of spontaneous activity. Using this framework, I examine the ontogeny of spontaneous network activity in in vitro neuronal networks from various brain regions, recorded on both single and multi-well MEAs. I also develop a framework for classifying these recordings according to their network type, based on quantitative features of their activity patterns. Next, I take a multi-view approach to classifying neuronal cell types using both the morphological and electrophysiological features of cells. I show that a number of multi-view clustering algorithms can more reliably differentiate between neuronal cell types in two existing data sets, compared to single-view clustering techniques applied to either the morphological or electrophysiological ‘view’ of the data, or a concatenation of the two views. To close, I examine the properties of the cell types identified by these methods.

Base teórica: o tratamento do carcinoma endometrial é feito através do estadiamento cirúrgico, que envolve histerectomia com salpingo-oforectomia bilateral e linfadenectomia pélvica e para-aórtica. Questiona-se o benefício da linfadenectomia sistemática em todos os pacientes, já que o risco de disseminação linfática em tumores de baixo risco é pequeno e não há evidências de benefício terapêutico em sua realização. Desse modo, tentam-se encontrar modos de determinar, na avaliação pré-operatória, quais são os pacientes que poderão se beneficiar da linfadenectomia e aqueles que podem prescindir do procedimento. Objetivos: avaliar a concordância da avaliação anatomopatológica entre a amostra endometrial pré-operatória e a peça cirúrgica das pacientes submetidas a tratamento cirúrgico primário do carcinoma de endométrio, correlacionando com características das pacientes e das amostras da patologia. Métodos: foram incluídos pacientes submetidos a tratamento cirúrgico para carcinoma de endométrio que tinham diagnóstico pré-operatório através de amostragem endometrial. Os prontuários foram revisados e as amostras disponíveis na instituição foram procuradas para posterior releitura por dois patologistas cegados para as demais informações anatomopatológicas. Resultados: foram incluídos 166 pacientes, com uma idade média de 64,6 anos. Das biópsias, 118 eram tumores endometrioides, 38 não-endometrioides e as demais, hiperplasia. As taxas de concordância foram de 93,2% para tumores endometrioides e 68,9% para não-endometrioides, com um índice kappa (k) de 0,73 para o tipo histológico. O grau tumoral distribui-se na amostra como G1 em 37,1%, G2 em 35,7% e G3 em 27,1%, com uma taxa de concordância de 61,5%, 56% e 78,9%, respectivamente, e k=0,46. Dos tumores G1, somente 1,9% teve upgrade para G3, em comparação com 16% das lesões G2. Não houve diferença estatística na taxa de concordância do tipo histológico e grau tumoral em função do local de execução da biópsia, método de amostragem e intervalo biópsia-cirurgia. Biópsias com pés > 3g tiveram uma concordância do grau tumoral significativamente melhor (p=0,040). Amostras de 105 pacientes estavam disponíveis no HCPA e foram reavaliadas por dois patologistas, com uma taxa de concordância interobservador geral de 73,3% (k=0,58) para o tipo histológico e 57,9% (k=0,54) para o grau tumoral. Conclusão: a acurácia da biópsia pré-operatória em predizer as características da peça cirúrgica não é ideal. Deve-se ter cuidado ao utilizar essa informação para determinar a extensão da cirurgia a ser realizada, sob risco de ser realizado subestadiamento. Estas baixas taxas de concordância correlacionam-se também com as baixas taxas de concordância interobservador. Novos sistemas de graduação e equipes de especialistas são possibilidades para melhorar esta questão. / Background: endometrial carcinoma treatment is based on surgical staging, including hysterectomy with bilateral salpingo-oophorectomy and pelvic and paraortic lymphadenectomy. The benefits of systematic lymphadenectomy in all patients have been questioned, since the risk of dissemination in low risk tumors is small and there is no evidence of benefits in its execution. Thereby, researches are looking for ways to determine, by preoperative evaluation, which patient will benefit from full staging and those who can do without the procedure. Objectives: evaluate the agreement between the preoperative endometrial samples and the surgical specimens in endometrial carcinoma, correlating it with characteristics of the samples and patients included, and evaluate the interobserver agreement of the preoperative biopsy. Methods: patients submitted to surgery as primary treatment for endometrial carcinoma at HCPA with a preoperative endometrial sampling were included. Their medical charts were reviewed. The available samples of the preoperative biopsies were recollected for reanalyzes by two pathologists. Inadequate transcriptions of the biopsy report were excluded. Results: we included 166 patients, with a mean age of 64.6 years. Of the biopsies, 118 were endometrioid, 38 were non-endometrioid and the remaining, hyperplasia. The agreement rates were 93.2% for endometrioid tumors and 68.9% for non-endometrioid, with a kappa index of 0.73 for the tumor cell type. The tumor FIGO grade distributed as G1 in 37.1%, G2 in 35.7% and G3 in 27.1%, with an agreement rate of 61.5%, 56% and 78.9%, respectively. The general kappa index for FIGO grading was 0.46. Of the G1 tumors, only 1.9% upgraded to G3, while 16% of the G2 lesions upgraded. There was no statistical difference in the agreement rates of tumor cell type and FIGO grading in function of place of biopsy execution, method of endometrial sampling and biopsy-surgery interval. Biopsies weighing more than 3g had a significantly better agreement in FIGO grading (p=0.040). Samples of 105 were available at HCPA and were reevaluated by 2 pathologists, with a general interobserver agreement 73.3/% (k=0.58) for tumor cell type and 57.9% (k=0.54) for grading. Conclusion: the accuracy of the preoperative biopsy in predicting the definite surgical characteristics it is not ideal. Caution must be taken when using this information to determine the surgical extension, due to the risk of under staging. These low rates of agreement are correlated with the low interobserver agreement. New grading systems and specialists teams are possible ways of improving this issue.

Base teórica: o tratamento do carcinoma endometrial é feito através do estadiamento cirúrgico, que envolve histerectomia com salpingo-oforectomia bilateral e linfadenectomia pélvica e para-aórtica. Questiona-se o benefício da linfadenectomia sistemática em todos os pacientes, já que o risco de disseminação linfática em tumores de baixo risco é pequeno e não há evidências de benefício terapêutico em sua realização. Desse modo, tentam-se encontrar modos de determinar, na avaliação pré-operatória, quais são os pacientes que poderão se beneficiar da linfadenectomia e aqueles que podem prescindir do procedimento. Objetivos: avaliar a concordância da avaliação anatomopatológica entre a amostra endometrial pré-operatória e a peça cirúrgica das pacientes submetidas a tratamento cirúrgico primário do carcinoma de endométrio, correlacionando com características das pacientes e das amostras da patologia. Métodos: foram incluídos pacientes submetidos a tratamento cirúrgico para carcinoma de endométrio que tinham diagnóstico pré-operatório através de amostragem endometrial. Os prontuários foram revisados e as amostras disponíveis na instituição foram procuradas para posterior releitura por dois patologistas cegados para as demais informações anatomopatológicas. Resultados: foram incluídos 166 pacientes, com uma idade média de 64,6 anos. Das biópsias, 118 eram tumores endometrioides, 38 não-endometrioides e as demais, hiperplasia. As taxas de concordância foram de 93,2% para tumores endometrioides e 68,9% para não-endometrioides, com um índice kappa (k) de 0,73 para o tipo histológico. O grau tumoral distribui-se na amostra como G1 em 37,1%, G2 em 35,7% e G3 em 27,1%, com uma taxa de concordância de 61,5%, 56% e 78,9%, respectivamente, e k=0,46. Dos tumores G1, somente 1,9% teve upgrade para G3, em comparação com 16% das lesões G2. Não houve diferença estatística na taxa de concordância do tipo histológico e grau tumoral em função do local de execução da biópsia, método de amostragem e intervalo biópsia-cirurgia. Biópsias com pés > 3g tiveram uma concordância do grau tumoral significativamente melhor (p=0,040). Amostras de 105 pacientes estavam disponíveis no HCPA e foram reavaliadas por dois patologistas, com uma taxa de concordância interobservador geral de 73,3% (k=0,58) para o tipo histológico e 57,9% (k=0,54) para o grau tumoral. Conclusão: a acurácia da biópsia pré-operatória em predizer as características da peça cirúrgica não é ideal. Deve-se ter cuidado ao utilizar essa informação para determinar a extensão da cirurgia a ser realizada, sob risco de ser realizado subestadiamento. Estas baixas taxas de concordância correlacionam-se também com as baixas taxas de concordância interobservador. Novos sistemas de graduação e equipes de especialistas são possibilidades para melhorar esta questão. / Background: endometrial carcinoma treatment is based on surgical staging, including hysterectomy with bilateral salpingo-oophorectomy and pelvic and paraortic lymphadenectomy. The benefits of systematic lymphadenectomy in all patients have been questioned, since the risk of dissemination in low risk tumors is small and there is no evidence of benefits in its execution. Thereby, researches are looking for ways to determine, by preoperative evaluation, which patient will benefit from full staging and those who can do without the procedure. Objectives: evaluate the agreement between the preoperative endometrial samples and the surgical specimens in endometrial carcinoma, correlating it with characteristics of the samples and patients included, and evaluate the interobserver agreement of the preoperative biopsy. Methods: patients submitted to surgery as primary treatment for endometrial carcinoma at HCPA with a preoperative endometrial sampling were included. Their medical charts were reviewed. The available samples of the preoperative biopsies were recollected for reanalyzes by two pathologists. Inadequate transcriptions of the biopsy report were excluded. Results: we included 166 patients, with a mean age of 64.6 years. Of the biopsies, 118 were endometrioid, 38 were non-endometrioid and the remaining, hyperplasia. The agreement rates were 93.2% for endometrioid tumors and 68.9% for non-endometrioid, with a kappa index of 0.73 for the tumor cell type. The tumor FIGO grade distributed as G1 in 37.1%, G2 in 35.7% and G3 in 27.1%, with an agreement rate of 61.5%, 56% and 78.9%, respectively. The general kappa index for FIGO grading was 0.46. Of the G1 tumors, only 1.9% upgraded to G3, while 16% of the G2 lesions upgraded. There was no statistical difference in the agreement rates of tumor cell type and FIGO grading in function of place of biopsy execution, method of endometrial sampling and biopsy-surgery interval. Biopsies weighing more than 3g had a significantly better agreement in FIGO grading (p=0.040). Samples of 105 were available at HCPA and were reevaluated by 2 pathologists, with a general interobserver agreement 73.3/% (k=0.58) for tumor cell type and 57.9% (k=0.54) for grading. Conclusion: the accuracy of the preoperative biopsy in predicting the definite surgical characteristics it is not ideal. Caution must be taken when using this information to determine the surgical extension, due to the risk of under staging. These low rates of agreement are correlated with the low interobserver agreement. New grading systems and specialists teams are possible ways of improving this issue.

Base teórica: o tratamento do carcinoma endometrial é feito através do estadiamento cirúrgico, que envolve histerectomia com salpingo-oforectomia bilateral e linfadenectomia pélvica e para-aórtica. Questiona-se o benefício da linfadenectomia sistemática em todos os pacientes, já que o risco de disseminação linfática em tumores de baixo risco é pequeno e não há evidências de benefício terapêutico em sua realização. Desse modo, tentam-se encontrar modos de determinar, na avaliação pré-operatória, quais são os pacientes que poderão se beneficiar da linfadenectomia e aqueles que podem prescindir do procedimento. Objetivos: avaliar a concordância da avaliação anatomopatológica entre a amostra endometrial pré-operatória e a peça cirúrgica das pacientes submetidas a tratamento cirúrgico primário do carcinoma de endométrio, correlacionando com características das pacientes e das amostras da patologia. Métodos: foram incluídos pacientes submetidos a tratamento cirúrgico para carcinoma de endométrio que tinham diagnóstico pré-operatório através de amostragem endometrial. Os prontuários foram revisados e as amostras disponíveis na instituição foram procuradas para posterior releitura por dois patologistas cegados para as demais informações anatomopatológicas. Resultados: foram incluídos 166 pacientes, com uma idade média de 64,6 anos. Das biópsias, 118 eram tumores endometrioides, 38 não-endometrioides e as demais, hiperplasia. As taxas de concordância foram de 93,2% para tumores endometrioides e 68,9% para não-endometrioides, com um índice kappa (k) de 0,73 para o tipo histológico. O grau tumoral distribui-se na amostra como G1 em 37,1%, G2 em 35,7% e G3 em 27,1%, com uma taxa de concordância de 61,5%, 56% e 78,9%, respectivamente, e k=0,46. Dos tumores G1, somente 1,9% teve upgrade para G3, em comparação com 16% das lesões G2. Não houve diferença estatística na taxa de concordância do tipo histológico e grau tumoral em função do local de execução da biópsia, método de amostragem e intervalo biópsia-cirurgia. Biópsias com pés > 3g tiveram uma concordância do grau tumoral significativamente melhor (p=0,040). Amostras de 105 pacientes estavam disponíveis no HCPA e foram reavaliadas por dois patologistas, com uma taxa de concordância interobservador geral de 73,3% (k=0,58) para o tipo histológico e 57,9% (k=0,54) para o grau tumoral. Conclusão: a acurácia da biópsia pré-operatória em predizer as características da peça cirúrgica não é ideal. Deve-se ter cuidado ao utilizar essa informação para determinar a extensão da cirurgia a ser realizada, sob risco de ser realizado subestadiamento. Estas baixas taxas de concordância correlacionam-se também com as baixas taxas de concordância interobservador. Novos sistemas de graduação e equipes de especialistas são possibilidades para melhorar esta questão. / Background: endometrial carcinoma treatment is based on surgical staging, including hysterectomy with bilateral salpingo-oophorectomy and pelvic and paraortic lymphadenectomy. The benefits of systematic lymphadenectomy in all patients have been questioned, since the risk of dissemination in low risk tumors is small and there is no evidence of benefits in its execution. Thereby, researches are looking for ways to determine, by preoperative evaluation, which patient will benefit from full staging and those who can do without the procedure. Objectives: evaluate the agreement between the preoperative endometrial samples and the surgical specimens in endometrial carcinoma, correlating it with characteristics of the samples and patients included, and evaluate the interobserver agreement of the preoperative biopsy. Methods: patients submitted to surgery as primary treatment for endometrial carcinoma at HCPA with a preoperative endometrial sampling were included. Their medical charts were reviewed. The available samples of the preoperative biopsies were recollected for reanalyzes by two pathologists. Inadequate transcriptions of the biopsy report were excluded. Results: we included 166 patients, with a mean age of 64.6 years. Of the biopsies, 118 were endometrioid, 38 were non-endometrioid and the remaining, hyperplasia. The agreement rates were 93.2% for endometrioid tumors and 68.9% for non-endometrioid, with a kappa index of 0.73 for the tumor cell type. The tumor FIGO grade distributed as G1 in 37.1%, G2 in 35.7% and G3 in 27.1%, with an agreement rate of 61.5%, 56% and 78.9%, respectively. The general kappa index for FIGO grading was 0.46. Of the G1 tumors, only 1.9% upgraded to G3, while 16% of the G2 lesions upgraded. There was no statistical difference in the agreement rates of tumor cell type and FIGO grading in function of place of biopsy execution, method of endometrial sampling and biopsy-surgery interval. Biopsies weighing more than 3g had a significantly better agreement in FIGO grading (p=0.040). Samples of 105 were available at HCPA and were reevaluated by 2 pathologists, with a general interobserver agreement 73.3/% (k=0.58) for tumor cell type and 57.9% (k=0.54) for grading. Conclusion: the accuracy of the preoperative biopsy in predicting the definite surgical characteristics it is not ideal. Caution must be taken when using this information to determine the surgical extension, due to the risk of under staging. These low rates of agreement are correlated with the low interobserver agreement. New grading systems and specialists teams are possible ways of improving this issue.

Recent research in bioinformatics has presented multiple cell type identification meth- dologies using single cell RNA sequence data (scRNA-seq). However, a consensus on which cell typing methodology consistently demonstrates superior performance remains absent. Additionally, very few studies approach cell type identification through a semi- supervised learning study, whereby the information in unlabeled data is leveraged to train an enhanced classifier. This paper presents cell annotation methodologies through self- learning and graph-based semi-supervised learning, in both raw count scRNA-seq data as well as in a latent embedding. I find that a self-learning framework enhances perfor- mance compared to a solely supervised learning classifier. Additionally, modelling on the latent data representations consistently outperforms modelling on the original data. The results show an overall accuracy of 96.12%, whereas additional models achieve an average precision rate of 95.12% and an average recall rate of 94.40%. The semi-supervised learn- ing approaches in this thesis compare favourable to scANVI in terms of accuracy, average precision rate, average recall rate and average f1-score. Moreover, results for alternative scenarios, in which cell types among training and test data do not perfectly overlap, are reported in this thesis.