Leukemi är en form av cancer som kan vara en dödlig sjukdom. För att rehabilitera och behandla sjukdomen krävs det en korrekt och tidig diagnostisering. För att minska väntetiden för testresultaten har de ordinära metoderna transformerats till automatiserade datorverktyg som kan analyser och diagnostisera symtom.I detta arbete, utfördes det en komparativ studie. Det man jämförde var två olika metoder som detekterar leukemia. Den ena metoden är en genetisk sekvenserings metod som är en binär klassificering och den andra metoden en bildbehandlings metod som är en fler-klassad klassificeringsmodell. Modellerna hade olika inmatningsvärden, däremot använde sig de båda av Convolutional neural network (CNN) som nätverksarkitektur och fördelade datavärdena med en 3-way cross-validation teknik. Utvärderings metoderna för att analysera resultaten är learning curves, confusion matrix och klassifikation rapport. Resultaten visade att den genetiska sekvenserings metoden hade fler antal värden som var korrekt förutsagda med 98 % noggrannhet. Den presterade bättre än bildbehandlings metoden som hade värde på 81% noggrannhet. Storlek på de olika datauppsättningar kan vara en orsak till algoritmernas olika testresultat. / Leukemia is a form of cancer that can be a fatal disease, and to rehabilitate and treat it requires a correct and early diagnosis. Standard methods have transformed into automated computer tools for analyzing, diagnosing, and predicting symptoms.In this work, a comparison study was performed by comparing two different leukemia detection methods. The methods were a genomic sequencing method, which is a binary classification model and a multi-class classification model, which was an images-processing method. The methods had different input values. However, both of them used a Convolutional neural network (CNN) as network architecture. They also split their datasets ​​using 3-way cross-validation. The evaluation methods for analyzing the results were learning curves, confusion matrix, and classification report. The results showed that the genome model had better performance and had several numbers of values ​​that were correctly predicted with a total accuracy of 98%. This value was compared to the image processing method results that have a value of 81% total accuracy. The size of the different data sets can be a cause of the different test results of the algorithms.

Le cannibalisme d’une cellule vivante par une autre cellule vivante représente une nouvelle modalité de mort cellulaire non autonome. Mes travaux de thèse ont permis d’identifier et de caractériser les acteurs impliqués et d’apprécier l’influence de ce processus sur le devenir de la cellule cannibale. Nous avons ainsi révélé que l’activation d’une signalisation cellulaire impliquée dans la régulation du cycle cellulaire va causer une libération d’ATP qui stimulera les récepteurs purinergiques P2Y2 de la cellule de manière autocrine. Cette étape sera suivie d’une augmentation de l’exposition de la protéine d’adhérence E-cadhérine à la membrane plasmique et de réarrangements du cytosquelettes médiés par la kinase ROCK, et permettra ainsi à une cellule vivante de cannibaliser une autre cellule vivante. Ce phénomène aussi connu sous le nom de « cellule dans une cellule » est fréquemment observé dans les biopsies tumorales. De plus, nous révélons au cours de ces travaux la capacité des cellules internalisées à être éliminées par un processus qui implique la protéine de l’autophagie ATG5 et les protéines pro-apoptotiques BAK et BAX. Ce processus est associé au déclenchement d’une instabilité génétique et d’un stress oxydatif au niveau des cellules cannibales et va déclencher la sénescence de ces cellules que nous avons appelé « entescence ». Cette nouvelle modalité d’induction de la sénescence participe à la suppression des tumeurs in vivo et semble prédire la réponse des patients aux traitements néoadjuvants anticancéreux. À l’opposé, l’échappement à l’entescence favorise la progression tumorale et est associé à une mauvaise réponse des patients aux traitements. L’ensemble de ces travaux met en lumière l’existence d’une nouvelle modalité d’induction de la sénescence cellulaire qui survient à la suite du cannibalisme cellulaire. Une meilleure compréhension des mécanismes impliqués dans son déclenchement et son exécution pourrait selon nous participer au développement de nouvelles approches thérapeutiques afin de lutter contre le cancer. / Cannibalism of live cells by other live cells is a new modality of non-autonomous cell death. This investigation led to the characterization of the molecular mechanisms implicated as well as the identification of the consequences of this process on the fate of the cannibal cell.We revealed that the activation of a signaling pathway involved in the regulation of the cell cycle can trigger a release of ATP that will stimulate the activity of the P2Y2 purinergic receptor in an autocrine manner. These events will lead to the increase of E-cadherin membrane exposition and change the organisation of the cytoskeleton in a ROCK-dependent manner, allowing this live cell to eat another live cell. This process called "cell in cell structure" is frequently observed in tumoral biopsies. Then, we revealed that the internalized cell will be eliminated by a process dependent on the autophagy protein ATG5 and the pro-apoptotic proteins BAX and BAK. These events are associated to the triggering of genomic instability and an oxidative stress in the cannibal cell leading these cells to a new senescence program that we called "entescence".This new senescence program seems to be a tumor suppressor mechanisms in vivo and is correlated to a better response of patient to neoadjuvant anticancer treatments. Moreover, escaping entescence seems to favor tumor growth and is associated to a bad response to anticancer treatments.Taken together, these results highlight the existence of a new senescence program that is initiated by cellular cannibalism. A better understanding of the molecular mechanisms regulating its initiation and its execution may lead to develop new innovative anticancer therapeutical approaches.

L’effectif génétique réduit des races bovines entraîne une augmentation de consanguinité de l’ordre de 1% par génération et une forte dérive génétique. Cette évolution favorise l’émergence régulière d’anomalies génétiques récessives dans les populations, qu’elles soient de races laitières ou allaitantes. En France, l’Observatoire National des Anomalies Bovines (ONAB) a été créé dans le but de détecter et contrôler ces anomalies émergentes. Cependant, la détection par les observatoires sous-entend d’une part une diffusion large de l’allèle défavorable dans la population et d’autre part que l’anomalie présente un phénotype avec un tableau clinique spécifique permettant une déclaration du phénotype à l’ONAB. L’impact des anomalies génétiques est donc encore largement sous-estimé. Toutefois, le développement des technologies de génotypage et de séquençage de génomes, associés à l’ensemble des informations disponibles permet une détection efficace des mutations causales. Ainsi, l’objectif de cette thèse a été d’utiliser l’ensemble des données disponibles (phénotypes, génotypes, séquences, annotations fonctionnelles…) pour identifier et valider des mutations délétères ségrégant dans races bovines laitières et allaitantes françaises. Nous avons exploré différentes stratégies classiques - cartographies par homozygotie, haplotypes en déficit en homozygotes - qui gagnent en efficacité grâce aux données de séquence de génome complet (WGS). Nous avons également mis en place des approches alternatives de génétique inverse, basées sur l’exploitation des données WGS française et du consortium 1000 bull genomes.Les travaux réalisé ont permis d’identifier les mutations causales associées à deux syndromes récessifs rapportés à l’ONAB : l’épidermolyse bulleuse jonctionnelle en race Charolaise (ITGB4, g.chr19: g.56488278_56493087del) et d’épilepsie idiopathique (MTCL1, g.chr24:41661691 G>A) récemment émergée dans la race Parthenaise. Nous avons également démontré la forte association entre l’haplotype MH1 et un polymorphisme affectant le gène PFAS (Chr19:g.28511199C>T ; p.R1205C). Par les stratégies de génétique inverse, nous avons également identifié une mutation probablement responsable de mortalité embryonnaire en race Normande affectant le gène CAD (Chr11:g72399397, p.Y452C) ainsi qu’une mutation affectant le gène RP1 (Chr14:g.23995411_23995412insA, p. R791KfsX13) responsable d’une dégénérescence progressive de la rétine et qui ségrége à forte fréquence en race Normande mais aussi dans d’autres races bovines européennes. Ces études encore en cours, fournissent également un inventaire des variants génétiques potentiellement délétères dont la caractérisation de l’effet sur le phénotype pourra être explorée. Enfin, l’identification de ces anomalies et des mutations délétères responsables ont abouti à la mise à disposition de tests de diagnostic efficaces pour permettre une contre-sélection raisonnée de ces variants délétères dans les populations bovines. / The reduced genetic size of cattle breeds leads to an increase in inbreeding of nearly 1% per generation and a strong genetic drift. This evolution favors regular emergence of recessive genetic abnormalities in dairy and beef cattle breeds. In France, the National Observatory of Bovine Anomalies (ONAB) was created with the aim to detect and control these new emergences. However, detection by the observatories implies a broad diffusion of the deleterious allele in the population and a phenotype with specific clinical features, allowing reporting to ONAB. Therefore the impact of genetic abnormalities is still largely underestimated. However, development of genotyping and genome sequencing technologies, coupled with all available information, makes possible effective detection of causal mutations. Thus, the aim of this thesis was to use all the available data (phenotypes, genotypes, sequences and functional annotations) to identify and validate deleterious mutations segregating in French dairy and beef cattle breeds. We used homozygosity mapping, and study of haplotypes with deficit in homozygous, two strategies boosted by using whole genome sequence (WGS) data. We also explored alternative reverse genetics strategies, based on data mining of French and 1000 bull genomes consortium WGS data.In these different studies, we identified the causal mutations associated with two described recessive syndromes: epidermolysis bullosa junctional in Charolais race (ITGB4, chr19: g.56488278_56493087del) and idiopathic epilepsy (MTCL1, chr24:g.41661691G>A) recently emerged in the Parthenaise breed. We also demonstrated a strong association between the embryonic lethal mutation MH1 and a polymorphism affecting the PFAS gene (chr19: g.28511199C> T ; p.R1205C) in Montbeliarde cattle. With reverse genetic strategies, we identified another mutation, which affects the CAD gene (chr11: g72399397, p.Y452C), likely responsible for embryonic mortality in Normande cattle, and a frameshift mutation in RP1 (chr14:g.23995411_23995412insA, p. R791KfsX13) responsible for progressive retinal degeneration segregating with a high frequency in the Normande breed but also in other European cattle breeds. These studies, still in progress, provide an inventory of potentially deleterious genetic variants, whose characterization could be explored in the future. At last, identification of mutations responsible for these genetic abnormalities provided or will provide diagnostic tests and allow efficient counter selection of these variants in French beef and dairy cattle populations.

En sélection génomique, des marqueurs de l’ADN sont utilisés pour l’évaluation des grandes races laitières. La plupart des méthodes d’évaluation génomique actuelles utilisent des SNP, bien que l’utilisation d’haplotypes de SNP apporte un plus grand polymorphisme. Il n’y avait pas d’évaluation génomique en place en 2014 pour les races régionales (Abondance, Tarentaise, Vosgienne), plaçant ces races en position de faiblesse.Notre objectif principal a été de mesurer l’intérêt de l’utilisation d’haplotypes en évaluation génomique, y compris à partir d’une population d’apprentissage multiraciale. Nous avons montré que les haplotypes conduisent à de meilleurs résultats que les SNP et que la fréquence des allèles et l’étendu du déséquilibre de liaison sont importants pour une construction optimale des haplotypes. Nous avons développé deux critères incorporant ces informations qui améliorent la précision des évaluations tout en réduisant le nombre de marqueurs utilisés.Depuis 2015, un de ces critères a été inclus dans les évaluations génomiques officielles en France. Notre approche a donné dans les races régionales une précision similaire à celle obtenue après testage sur descendance. Une évaluation génomique de routine est en place pour 3 races régionales en France depuis Juin 2016. L’utilisation d’une puce Haute Densité n’a pas amélioré sa précision, alors qu’une population d’apprentissage multiraciale a été bénéfique uniquement pour certaines races. Le génotypage des nouvelle femelles a augmenté la précision de la sélection mais l’inclusion de mutations candidates détectées dans les grandes races laitières n’a conduit qu’à une légère amélioration chez les races régionales. / In genomic selection, DNA marker information is exploited for evaluation purposes in large dairy cattle breeds. Most of the current genomic evaluation methods rely today on SNP information, although haplotypes are expected to perform better due to their higher polymorphism. In 2014, genomic evaluation had not yet been implemented in regional breeds (Abondance, Tarentaise, Vosgienne), resulting in economic weaknesses for these breeds.Our aim was to assess the use of haplotypes in genomic evaluation with focus on their performance in combination with multi-breed reference populations. We found that haplotypes outperformed individual SNP markers for genomic evaluation. We also showed that information on haplotype allele frequency and on linkage pattern are relevant to select haplotypes for evaluation purposes. Our haplotype selection criteria also allowed a significant reduction of the number of markers used for genomic prediction.One of these criteria was incorporated into the French routine genomic evaluation in 2015. The performance of such an evaluation was then assessed in four regional breeds, leading to similar or higher accuracies than current progeny testing. Consequently, routine genomic evaluation was implemented in these breeds in 2016. The use of high density genotypes did not improve the performance of genomic evaluation in these breeds, while multi-breed training populations were beneficial only in some of them. Additional genotyped females led to notable increases in selection accuracies. Inclusion of candidate mutations identified in large breeds led to only minor improvements in regional breeds.

Genomes of higher plants comprise a large proportion of repetitive DNAs, where one major class is satellite DNA. Satellite DNA is organized in tandem arrays of basic repeating units, which often occurs in heterochromatin of centromeric/pericentromeric and intercalary as well as subtelomeric regions. Besides these typical satellite repeats, there are also non-typical satellite DNAs, which are organized in short tandem arrays and integrated into a transposable element. The chromosomal localization of non-typical satellites is not in large regions of heterochromatin, but tend to be dispersed along chromosomes. This thesis describes the identification of the major repeat classes including major satellite content in six beet and related species. The focus was on identification and characterization of new satellite families in the beet genomes. In this study, the information regarding repetitive DNA as well as satellite families fraction in six beet and related species was gained based on graph-based clustering of next generation sequenced short sequence reads. The repeat proportion of the six analyzed species ranges from 34.4% in C. quinoa to 65.6% in B. lomatogona, in which the portion of nearly 50% belongs to B. vulgaris, B. nana, P. procumbens, and P. patellaris. Among all classes of repetitive DNAs, LTR retrotransposons are the most abundant repeat type in all analyzed genomes, which is a common feature of higher plant genomes. The other repeat sequences are DNA transposons, rDNA, and satellite DNA with variable portions in different species. A set of satellite families in each species was analyzed in detail and reflects the relationship between six species. The closely related relationship between B. lomatogona and B. nana as well as between P. procumbens and P. patellaris is affirmed by seven and 13 satellite families shared between two species, respectively. Similarly, the closer relationship between B. vulgaris and two species B. lomatogona and B. nana than between B. vulgaris and two species P. procumbens and P. patellaris from the sister-genus Patellifolia is also confirmed. C. quinoa is a distantly related species and this is reflected by vastly different satellite content. Therefore, satellite DNA analysis might be a useful tool to trace species evolution. In the B. lomatogona genome, by the application of RepeatExplorer tool, six novel tandemly repeated DNA sequences were identified and designated BlSat1-BlSat6. The three typical satellite families BlSat1, BlSat5, and BlSat6 are organized in tandem arrays in large heterochromatic blocks. BlSat1 is mainly localized in the pericentric region of the chromosome 3, 5, 6, and 9, while BlSat5 is amplified in the pericentromeric region of the chromosome 3, 5, and 7. BlSat6 is a chromosome-specific satellite and is located in the subtelomeric region on the south arm of the chromosome 8. The other three satellite families BlSat2, BlSat3, and BlSat4 are characterized as non-typical satellite DNA because of their dispersed distribution along chromosomes. BlSat2 and BlSat3 are identified as a tandem repeat domain in Ogre/Tat retrotransposons. The occurrence of one or several short tandem arrays in a transposable element is a common phenomenon in both animals and plants. These short repeats are considered to be continuously evolving and eventually amplifying to new satellite families. Furthermore, the distribution of the six new satellite families in beet and related species was confirmed by comparative PCR, comparative Southern hybridization, and mapping of sequence reads from referent species against each satellite sequence. The BlSat1 and BlSat6 satellite families are specific for the genus Beta, while BlSat5 is only amplified in two sections Corollinae and Nanae of the genus Beta. BlSat4 is an ancient satellite family which exists in all tested species belonging to the genera Beta, Patellifolia, Chenopodium, and Spinacia, whereas BlSat2 and BlSat3 might have evolved before the separation of the genus Beta and Patellifolia but their sequences have been lost or heavily diverged during the species radiation. Comparison of two wild beet genomes P. procumbens and P. patellaris was performed aiming to address the open question whether P. patellaris is auto- or allotetraploid. The high similarity between these two genomes indicates their close relationship. However, the genetic difference between two genomes, in particular the molecular characteristics as well as the chromosomal localization of two satellite families PproSat1 and PpatSat1, might support a hypothesis that P. patellaris is allotetraploid species with a half of its chromosome set derived from P. procumbens. The results obtained in this work might provide comprehensive information of the repetitive classes as well as satellite families in the genomes of beets and related species. The results can be used as the species-specific and chromosome-specific markers in beet genome studies.

Embryonic stem (ES) cells have the ability to maintain pluripotency and self-renewal during in vitro maintenance, which is a key to their clinical applications. ES cells are a model in developmental biology studies due to their potential to differentiate in vitro. Understanding critical pathways of pluripotency, self-renewal, and differentiation during early embryonic development is important for the evaluation of the therapeutic potential of ES cells because of their ability for tumor transformation due to genetic and epigenetic instability acquired during in vitro culture maintenance. Single tandem repeats are sequences of DNA that have been implicated in the deregulation of gene expression in different human conditions. Understanding the origin of repetitive sequence instability and functions in the genome allow characterization of early genomic instability signals in ES cell pluripotency, differentiation, and tumor transformation pathways. The hypothesis of this study was that genetic stability, in repetitive sequences, located near embryonic developmental genes is responsible for pluripotency, self-renewal, differentiation, and chromatin assembly and could be a signal for adaptation, differentiation, or transformation of ES cells in vitro. Our result showed instability in specific repetitive sequences which increased during ES cell passages and embryoid body differentiation in vitro. ES cells displayed significant mean frequencies of genomic instability in repetitive regions that lead to ES cells pluripotency, self-renewal maintenance, or cell lineage specialization. The present study reports potentially biomarkers for identifying accumulation of genomic instability in specific genes that may contributes to adaptation of ES cells and could be the switch that initiates early ES cell lineage commitment in vitro. Determining genetic and epigenetic modifications, including single tandem repeat instability, gene expression changes, and chromatin modifications, is essential for elucidating possible molecular mechanisms of genomic instability and determining novel molecular characterization for diagnostic purposes to ensure ES cell stability and integrity that could potentially lead to use of ES cell derivatives that could then be a safe source needed for regenerative medicine applications