After reviewing Multifactor Dimensionality Reduction(MDR) and its extensions, an approach to obtain P(larger than 1) risk scores is proposed to predict the continuous outcome for each subject. We study the mean square error(MSE) of dimensionality reduced models fitted with sets of 2 risk scores and investigate the MSE for several special cases of the covariance matrix. A methodology is proposed to select a best set of P risk scores when P is specified a priori. Simulation studies based on true models of different dimensions(larger than 3) demonstrate that the selected set of P(larger than 1) risk scores outperforms the single aggregated risk score generated in AQMDR and illustrate that our methodology can determine a best set of P risk scores effectively. With different assumptions on the dimension of the true model, we considered the preferable set of risk scores from the best set of two risk scores and the best set of three risk scores. Further, we present a methodology to access a set of P risk scores when P is not given a priori. The expressions of asymptotic estimated mean square error of prediction(MSPE) are derived for a 1-dimensional model and 2-dimensional model. In the last main chapter, we apply the methodology of selecting a best set of risk scores where P has been specified a priori to Alzheimer’s Disease data and achieve a set of 2 risk scores and a set of three risk scores for each subject to predict measurements on biomarkers that are crucially involved in Alzheimer’s Disease.

The missing heritability in genome-wide association studies (GWAS) is an intriguing open scientific problem which has attracted great recent interest. The interaction effects among risk factors, both genetic and environmental, are hypothesized to be one of the main missing heritability sources. Moreover, detection of multilocus interaction effect may also have great implications for revealing disease/biological mechanisms, for accurate risk prediction, personalized clinical management, and targeted drug design. However, current analysis of GWAS largely ignores interaction effects, partly due to the lack of tools that meet the statistical and computational challenges posed by taking into account interaction effects. Here, we propose a novel statistically-based framework (Significant Conditional Association) for systematically exploring, assessing significance, and detecting interaction effect. Further, our SCA work has also revealed new theoretical results and insights on interaction detection, as well as theoretical performance bounds. Using in silico data, we show that the new approach has detection power significantly better than that of peer methods, while controlling the running time within a permissible range. More importantly, we applied our methods on several real data sets, confirming well-validated interactions with more convincing evidence (generating smaller p-values and requiring fewer samples) than those obtained through conventional methods, eliminating inconsistent results in the original reports, and observing novel discoveries that are otherwise undetectable. The proposed methods provide a useful tool to mine new knowledge from existing GWAS and generate new hypotheses for further research. Microarray gene expression studies provide new opportunities for the molecular characterization of heterogeneous diseases. Multiclass gene selection is an imperative task for identifying phenotype-associated mechanistic genes and achieving accurate diagnostic classification. Most existing multiclass gene selection methods heavily rely on the direct extension of two-class gene selection methods. However, simple extensions of binary discriminant analysis to multiclass gene selection are suboptimal and not well-matched to the unique characteristics of the multi-category classification problem. We report a simpler and yet more accurate strategy than previous works for multicategory classification of heterogeneous diseases. Our method selects the union of one-versus-everyone phenotypic up-regulated genes (OVEPUGs) and matches this gene selection with a one-versus-rest support vector machine. Our approach provides even-handed gene resources for discriminating both neighboring and well-separated classes, and intends to assure the statistical reproducibility and biological plausibility of the selected genes. We evaluated the fold changes of OVEPUGs and found that only a small number of high-ranked genes were required to achieve superior accuracy for multicategory classification. We tested the proposed OVEPUG method on six real microarray gene expression data sets (five public benchmarks and one in-house data set) and two simulation data sets, observing significantly improved performance with lower error rates, fewer marker genes, and higher performance sustainability, as compared to several widely-adopted gene selection and classification methods. / Ph. D.

Introduction – Animal studies suggested that NFKB1, SOCS3 and IKBKB genes could be involved in the association between overnutrition and obesity. This study aims to investigate interactions involving these genes and nutrition affecting obesity-related phenotypes. Methods – We used multifactor dimensionality reduction (MDR) and penalized logistic regression (PLR) to better detect gene/environment interactions in data from the Toronto Nutrigenomics and Health Study (n=1639) using dichotomized body mass index (BMI) and waist circumference (WC) as obesity-related phenotypes. Exposure variables included genotypes on 54 single nucleotide polymorphisms, dietary factors and ethnicity. Results – MDR identified interactions between SOCS3 rs6501199 and rs4969172, and IKBKB rs3747811 affecting BMI in whites; SOCS3 rs6501199 and NFKB1 rs1609798 affecting WC in whites; and SOCS3 rs4436839 and IKBKB rs3747811 affecting WC in South Asians. PLR found a main effect of SOCS3 rs12944581 on BMI among South Asians. Conclusion – MDR and PLR gave different results, but support some results from previous studies.

Cancer is a complex disease with diverse genetic changes that pose significant treatment challenges due to its heterogeneity. Identifying driver modules, which are crucial for cancer progression, has been aided by artificial intelligence (AI) techniques. However, existing approaches lack specificity, particularly for cancer types like lung cancer. This thesis addresses this gap by proposing a method that combines a gene-gene interaction network construction with AI-based clustering to identify distinct driver modules specific to lung cancer. The research aims to enhance our understanding of the disease by leveraging publicly available databases and large datasets using design science methodology. By mapping biological processes to genes and constructing a weighted gene-gene interaction network, correlations within gene clusters are identified. A clustering algorithm is applied to derive potential cancer-driver modules and pinpoint biologically relevant modules that contribute to the development of lung cancer. The results demonstrate the effectiveness and robustness of the clustering approach, with 110 unique and non-overlapping clusters identified, ranging in size from 4 to 10. These clusters surpass the evaluation requirements and exhibit significant relevance to critical pathways. The findings challenge previous assumptions about gene clusters and their significance in lung cancer, providing insights into the molecular underpinnings of the disease. The identified driver modules hold promise for influencing future approaches to diagnosis, prognosis, and treatment in the management of lung cancer. By expanding our understanding of the disease, this research paves the way for further investigations and potential clinical advancements.

Le mélanome cutané est un cancer des cellules de la peau (mélanocytes) qui se situe, en France, au 11e rang des cancers les plus fréquents. Sa mortalité reste élevée lorsqu’il est diagnostiqué à un stade tardif. Ce cancer résulte de nombreux facteurs génétiques, environnementaux et des interactions entre ces facteurs. La susceptibilité génétique à ce cancer recouvre un large spectre de variabilité génétique, depuis des mutations rares conférant un risque élevé jusqu’à des variants fréquents conférant un risque modeste. C’est dans le cadre de l’identification de variants fréquents liés à l’apparition du mélanome et à son pronostic que se situe mon travail de thèse. À ce jour, les études d’associations pangénomiques du mélanome ont identifié des variants fréquents à effets relativement modestes qui expliquent seulement une part de la composante génétique. Les variants fonctionnels au sein des régions identifiées sont le plus souvent inconnus. Les études pangénomiques ont eu principalement recours à des analyses simple-marqueur qui peuvent manquer de puissance pour détecter des variants ayant un effet individuel faible ou interagissant avec d’autres variants. L’objectif principal de ce travail de thèse a été de proposer des stratégies d’analyse multi-marqueurs pour identifier de nouveaux gènes impliqués dans le mélanome et pour caractériser des variants potentiellement fonctionnels au sein des régions du génome associées au mélanome.Pour identifier de nouveaux gènes associés au risque de mélanome et à un facteur pronostique de ce cancer (l’indice de Breslow), nous avons proposé une stratégie d’analyse multi-marqueurs qui intègre une analyse de pathways biologiques basée sur la méthode GSEA (Gene Set Enrichment Analysis) et une analyse d’interactions entre gènes au sein des pathways associés au mélanome. Ces analyses ont été menées dans deux études : l’étude française MELARISK et l’étude américaine du MD Anderson Cancer Center (MDACC), totalisant 2 980 cas et 3 823 témoins. Nous avons identifié une interaction entre les gènes, TERF1 et AFAP1L2, pour le risque de mélanome et une interaction entre les gènes, CDC42 et SCIN, pour l’indice de Breslow. Ces gènes sont particulièrement pertinents sur le plan biologique du fait de leur rôle dans la biologie des télomères pour la première paire de gènes et dans la dynamique des filaments d’actine pour la seconde paire. Afin d’identifier les variants potentiellement fonctionnels au sein des régions du génome mises en évidence par études pangénomiques, nous avons proposé une stratégie de cartographie fine qui repose principalement sur une méthode de régression pénalisée (méthode HyperLasso) appliquée à tous les variants de la région étudiée. Par l’analyse de la région 16q24 qui contient le gène MC1R dont les variants fonctionnels sont connus, nous avons montré que cette stratégie était capable d’identifier ces variants parmi de nombreux variants associés au mélanome dans cette région. Nous avons contribué à identifier cinq nouvelles régions du génome associées au mélanome par méta-analyse d’études pangénomiques réalisées au niveau mondial (43 000 sujets) puis mené une étude de cartographie fine de toutes les régions associées au mélanome, en se basant sur la stratégie proposée et validée dans la région 16q24. Les stratégies d’analyses multi-marqueurs proposées dans le cadre de ce travail de thèse ont permis d’identifier de nouveaux gènes associés au risque de mélanome et à un facteur pronostique de ce cancer et de caractériser les variants génétiques potentiellement fonctionnels au sein des régions du génome identifiées par études pangénomiques. / Cutaneous melanoma is a skin cancer developed from melanocytes. It is the 11th most common cancers in France. Mortality due to melanoma remains high when diagnosed at a late stage. This cancer results from many genetic, environmental factors and interactions between these factors. The genetic susceptibility to melanoma covers a broad spectrum of genetic variation, from rare mutations conferring high risk to common variants conferring low risk. My thesis was conducted in the framework of low-risk variants associated with melanoma occurrence and prognosis. To date, genome-wide association studies (GWAS) of melanoma have identified common variants with relatively modest effects which only explain a part of the genetic component of this cancer. Functional variants at the identified loci are mostly unknown. GWASs have been mainly conducted using single-marker analysis which may be underpowered to detect variants with small effect or interacting with each other. The main objective of this thesis was to propose multi-marker analysis strategies to identify novel genes involved in melanoma and to characterize potentially functional variants in chromosomal regions found associated with melanoma. To identify new genes associated with melanoma risk and a prognostic factor for this cancer (Breslow thickness), we proposed a multi-marker analysis strategy which integrates pathway analysis based on the GSEA (Gene Set Enrichment Analysis) method and gene-gene interaction analysis within melanoma-associated pathways. These analyses were conducted in two studies: the French MELARISK study and the North-American MD Anderson Cancer Center (MDACC) study, with a total of 2,980 cases and 3,823 controls. We identified gene-gene interactions between TERF1 and AFAP1L2 genes for melanoma risk and between CDC42 and SCIN genes for Breslow thickness. These genes are biologically relevant because of their role in telomere biology for the former gene pair and in actin dynamics for the latter pair. To identify potentially functional variants at loci identified by GWAS, we proposed a fine mapping strategy which is mainly based on a penalized regression approach (HyperLasso method) that can be applied to all variants of the region under study. By studying the 16q24 region which harbors the MC1R gene whose functional variants are known, we showed this strategy was able to identify those variants among many variants associated with melanoma in this region. We contributed to the identification of five novel regions associated with melanoma through a worldwide meta-analysis of melanoma GWASs (43,000 subjects) and conducted fine mapping of all melanoma-associated loci using the strategy we proposed and validated in the 16q24 region. The multi-marker strategies proposed in this work have allowed identifying new biologically relevant genes associated with risk of melanoma and a major melanoma prognostic factor and characterizing potentially functional genetic variants within regions identified by GWAS.

Worldwide, alcohol is the third leading risk factor for disease burden, while its harmful use leads to 2.5 million deaths every year. Alcohol dependence (AD) is a complex disease, with devastating effects on individuals, families, and society. It is estimated that 76.3 million people worldwide have suffered from alcohol use disorders (AUD), including alcohol abuse and AD. Alcohol withdrawal or alcohol withdrawal symptom (AWS) refers to a cluster of symptoms that may occur when a heavy drinker suddenly stops or significantly reduces their alcohol intake. These symptoms can start as early as 2 h after the last drink, persist for weeks, and range from mild anxiety and shakiness to severe complications, such as seizures and delirium tremens. Family, twin, and adoption studies have indicated that genetic and environmental factors and their interactions contribute to the development of AD and related phenotypes, with a heritability coefficient of more than 0.5 for AD. Whole-genome linkage and candidate gene association studies have successfully identified several chromosome regions and genes that are related to AD and AWS. Furthermore, gene expression analysis, epigenetic studies, and genome-wide association studies (GWAS) have provided regions and loci for AWS. This chapter reviews the recent findings in genetic studies of AWS.

La leucémie lymphoblastique aigüe (LLA) est une maladie génétique complexe. Malgré que cette maladie hématologique soit le cancer pédiatrique le plus fréquent, ses causes demeurent inconnues. Des études antérieures ont démontrées que le risque à la LLA chez l’enfant pourrait être influencé par des gènes agissant dans le métabolisme des xénobiotiques, dans le maintient de l’intégrité génomique et dans la réponse au stress oxydatif, ainsi que par des facteurs environnementaux. Au cours de mes études doctorales, j’ai tenté de disséquer davantage les bases génétiques de la LLA de l’enfant en postulant que la susceptibilité à cette maladie serait modulée, au moins en partie, par des variants génétiques agissant dans deux voies biologiques fondamentales : le point de contrôle G1/S du cycle cellulaire et la réparation des cassures double-brin de l’ADN. En utilisant une approche unique reposant sur l’analyse d’une cohorte cas-contrôles jumelée à une cohorte de trios enfants-parents, j’ai effectué une étude d’association de type gènes/voies biologiques candidats. Ainsi, j’ai évaluer le rôle de variants provenant de la séquence promotrice de 12 gènes du cycle cellulaire et de 7 gènes de la voie de réparation de l’ADN, dans la susceptibilité à la LLA. De tels polymorphismes dans la région promotrice (pSNPs) pourraient perturber la liaison de facteurs de transcription et mener à des différences dans les niveaux d’expression des gènes pouvant influencer le risque à la maladie. En combinant différentes méthodes analytiques, j’ai évalué le rôle de différents mécanismes génétiques dans le développement de la LLA chez l’enfant. J’ai tout d’abord étudié les associations avec gènes/variants indépendants, et des essaies fonctionnels ont été effectués afin d’évaluer l’impact des pSNPs sur la liaison de facteurs de transcription et l’activité promotrice allèle-spécifique. Ces analyses ont mené à quatre publications. Il est peu probable que ces gènes de susceptibilité agissent seuls; j’ai donc utilisé une approche intégrative afin d’explorer la possibilité que plusieurs variants d’une même voie biologique ou de voies connexes puissent moduler le risque de la maladie; ces travaux ont été soumis pour publication. En outre, le développement précoce de la LLA, voir même in utero, suggère que les parents, et plus particulièrement la mère, pourraient jouer un rôle important dans le développement de cette maladie chez l’enfant. Dans une étude par simulations, j’ai évalué la performance des méthodes d’analyse existantes de détecter des effets fœto-maternels sous un design hybride trios/cas-contrôles. J’ai également investigué l’impact des effets génétiques agissant via la mère sur la susceptibilité à la LLA. Cette étude, récemment publiée, fût la première à démontrer que le risque de la leucémie chez l’enfant peut être modulé par le génotype de sa mère. En conclusions, mes études doctorales ont permis d’identifier des nouveaux gènes de susceptibilité pour la LLA pédiatrique et de mettre en évidence le rôle du cycle cellulaire et de la voie de la réparation de l’ADN dans la leucémogenèse. À terme, ces travaux permettront de mieux comprendre les bases génétiques de la LLA, et conduiront au développement d’outils cliniques qui amélioreront la détection, le diagnostique et le traitement de la leucémie chez l’enfant. / Childhood acute lymphoblastic leukemia (ALL) is a complex and heterogeneous genetic disease. Although it is the most common pediatric cancer, its etiology remains poorly understood. Previous studies provided evidence that childhood ALL might originate through the collective contribution of different genes controlling the efficiency of carcinogen metabolism, the capacity of maintaining DNA integrity and the response to oxidative stress, as well as environmental factors. In my doctoral research project I attempted to further dissect the genetic intricacies underlying childhood ALL. I postulated that a child’s susceptibility to ALL may be influenced, in part, by functional sequence variation in genes encoding components of two core biologic pathways: G1/S cell cycle control and DNA double-strand break repair. Using a unique two-tiered study design consisting of both unrelated ALL cases and healthy controls, as well as case-parent trios, I performed a pathway-based candidate-gene association study to investigate the role of sequence variants in the promoter regions of 12 candidate cell cycle genes and 7 DNA repair genes, in modulating ALL risk among children. Polymorphisms in promoter regions (pSNPs) could perturb transcription factor binding and lead to differences in gene expression levels that in turn could modify the risk of disease. To better depict the complex genetic architecture of childhood ALL, I used multiple analytical approaches. First, individual genes/variants were tested for association with disease, while functional in vitro validation was performed to evaluate the impact of the pSNPs on differential transcription factor binding and allele-specific promoter activity. These analyses led to four published articles. Given that these genes are not likely to act alone to confer disease risk I used an integrative approach to explore the possibility that combinations of functionally relevant pSNPs among several components of the same or of interconnected pathways, could contribute to modified childhood ALL risk either through pathway-specific or epistatic effects; this work was recently submitted for publication. Finally, childhood ALL is thought to arise in utero suggesting that the parents, and in particular the mother, may play an important role in shaping disease susceptibility in their offspring. Using simulations, I investigated the performance of existing methods to test for maternal genotype associations using a case-parent trio/case-control hybrid design, and then assessed the impact of maternally-mediated genetic effects on ALL susceptibility among children. This published work was the first to show that the mother’s genotype can indeed influence the risk of leukemia in children, further corroborating the importance of considering parentally-mediated effects in the study of early-onset diseases. In conclusion, my doctoral work lead to the identification of novel genetic susceptibility loci for childhood ALL and provided evidence for the implication of the cell cycle control and DNA repair pathways in leukemogenesis. Better elucidation of the genetic mechanisms underlying the pathogenesis of ALL in children could be of great diagnostic value and provide data to help guide risk-directed therapy and improve disease management and outcome. Ultimately, this study brings us one step closer to unraveling the genetic architecture of childhood ALL and provides a stepping-stone towards disease prevention.

La leucémie lymphoblastique aigüe (LLA) est une maladie génétique complexe. Malgré que cette maladie hématologique soit le cancer pédiatrique le plus fréquent, ses causes demeurent inconnues. Des études antérieures ont démontrées que le risque à la LLA chez l’enfant pourrait être influencé par des gènes agissant dans le métabolisme des xénobiotiques, dans le maintient de l’intégrité génomique et dans la réponse au stress oxydatif, ainsi que par des facteurs environnementaux. Au cours de mes études doctorales, j’ai tenté de disséquer davantage les bases génétiques de la LLA de l’enfant en postulant que la susceptibilité à cette maladie serait modulée, au moins en partie, par des variants génétiques agissant dans deux voies biologiques fondamentales : le point de contrôle G1/S du cycle cellulaire et la réparation des cassures double-brin de l’ADN. En utilisant une approche unique reposant sur l’analyse d’une cohorte cas-contrôles jumelée à une cohorte de trios enfants-parents, j’ai effectué une étude d’association de type gènes/voies biologiques candidats. Ainsi, j’ai évaluer le rôle de variants provenant de la séquence promotrice de 12 gènes du cycle cellulaire et de 7 gènes de la voie de réparation de l’ADN, dans la susceptibilité à la LLA. De tels polymorphismes dans la région promotrice (pSNPs) pourraient perturber la liaison de facteurs de transcription et mener à des différences dans les niveaux d’expression des gènes pouvant influencer le risque à la maladie. En combinant différentes méthodes analytiques, j’ai évalué le rôle de différents mécanismes génétiques dans le développement de la LLA chez l’enfant. J’ai tout d’abord étudié les associations avec gènes/variants indépendants, et des essaies fonctionnels ont été effectués afin d’évaluer l’impact des pSNPs sur la liaison de facteurs de transcription et l’activité promotrice allèle-spécifique. Ces analyses ont mené à quatre publications. Il est peu probable que ces gènes de susceptibilité agissent seuls; j’ai donc utilisé une approche intégrative afin d’explorer la possibilité que plusieurs variants d’une même voie biologique ou de voies connexes puissent moduler le risque de la maladie; ces travaux ont été soumis pour publication. En outre, le développement précoce de la LLA, voir même in utero, suggère que les parents, et plus particulièrement la mère, pourraient jouer un rôle important dans le développement de cette maladie chez l’enfant. Dans une étude par simulations, j’ai évalué la performance des méthodes d’analyse existantes de détecter des effets fœto-maternels sous un design hybride trios/cas-contrôles. J’ai également investigué l’impact des effets génétiques agissant via la mère sur la susceptibilité à la LLA. Cette étude, récemment publiée, fût la première à démontrer que le risque de la leucémie chez l’enfant peut être modulé par le génotype de sa mère. En conclusions, mes études doctorales ont permis d’identifier des nouveaux gènes de susceptibilité pour la LLA pédiatrique et de mettre en évidence le rôle du cycle cellulaire et de la voie de la réparation de l’ADN dans la leucémogenèse. À terme, ces travaux permettront de mieux comprendre les bases génétiques de la LLA, et conduiront au développement d’outils cliniques qui amélioreront la détection, le diagnostique et le traitement de la leucémie chez l’enfant. / Childhood acute lymphoblastic leukemia (ALL) is a complex and heterogeneous genetic disease. Although it is the most common pediatric cancer, its etiology remains poorly understood. Previous studies provided evidence that childhood ALL might originate through the collective contribution of different genes controlling the efficiency of carcinogen metabolism, the capacity of maintaining DNA integrity and the response to oxidative stress, as well as environmental factors. In my doctoral research project I attempted to further dissect the genetic intricacies underlying childhood ALL. I postulated that a child’s susceptibility to ALL may be influenced, in part, by functional sequence variation in genes encoding components of two core biologic pathways: G1/S cell cycle control and DNA double-strand break repair. Using a unique two-tiered study design consisting of both unrelated ALL cases and healthy controls, as well as case-parent trios, I performed a pathway-based candidate-gene association study to investigate the role of sequence variants in the promoter regions of 12 candidate cell cycle genes and 7 DNA repair genes, in modulating ALL risk among children. Polymorphisms in promoter regions (pSNPs) could perturb transcription factor binding and lead to differences in gene expression levels that in turn could modify the risk of disease. To better depict the complex genetic architecture of childhood ALL, I used multiple analytical approaches. First, individual genes/variants were tested for association with disease, while functional in vitro validation was performed to evaluate the impact of the pSNPs on differential transcription factor binding and allele-specific promoter activity. These analyses led to four published articles. Given that these genes are not likely to act alone to confer disease risk I used an integrative approach to explore the possibility that combinations of functionally relevant pSNPs among several components of the same or of interconnected pathways, could contribute to modified childhood ALL risk either through pathway-specific or epistatic effects; this work was recently submitted for publication. Finally, childhood ALL is thought to arise in utero suggesting that the parents, and in particular the mother, may play an important role in shaping disease susceptibility in their offspring. Using simulations, I investigated the performance of existing methods to test for maternal genotype associations using a case-parent trio/case-control hybrid design, and then assessed the impact of maternally-mediated genetic effects on ALL susceptibility among children. This published work was the first to show that the mother’s genotype can indeed influence the risk of leukemia in children, further corroborating the importance of considering parentally-mediated effects in the study of early-onset diseases. In conclusion, my doctoral work lead to the identification of novel genetic susceptibility loci for childhood ALL and provided evidence for the implication of the cell cycle control and DNA repair pathways in leukemogenesis. Better elucidation of the genetic mechanisms underlying the pathogenesis of ALL in children could be of great diagnostic value and provide data to help guide risk-directed therapy and improve disease management and outcome. Ultimately, this study brings us one step closer to unraveling the genetic architecture of childhood ALL and provides a stepping-stone towards disease prevention.