The link between the genotype and the phenotype of an organism is immensely complex. Despite this it can, to a great extent, be captured using models that assume that gene variants combine their effects in an additive manner. This thesis explores aspects of genetics that cannot be fully captured using such additive models. Using experimental data from three different model organisms, I study two phenomena that fall outside of the additive paradigm: genetic interactions and genetic variance heterogeneity. Using the model plant Arabidopsis thaliana, we show how important biological insights can be reached by exploring loci that display genetic variance heterogeneity. In the first study, this approach identified alleles in the gene CMT2 associated with the climate at sampling locations, suggesting a role in climate adaption. These alleles affected the genome wide methylation pattern, and a complete knock down of this gene increased the plants heat tolerance. In the second study, we demonstrate how the observed genetic variance heterogeneity was the result of the partial linkage of many functional alleles near the gene MOT1, all contributing to Molybdenum levels in the leaves. Further, we explore genetic interactions using data from dogs and budding yeast (Saccharomyces cerevisiae). In the dog population, two interacting loci were associated with fructosamine levels, a biomarker used to monitor blood glucose. One of the loci displayed the pattern of a selective sweep in some of the studied breeds, suggesting that the interaction is important for the phenotypic breed-differences. In a cross between two strains of yeast, with the advantage of large population size and nearly equal allele frequencies, we identified large epistatic networks. The networks were largely centered on a number of hub-loci and altogether involved hundreds of genetic interactions. Most network hubs had the ability to either suppress or uncover the phenotypic effects of other loci. Many multi-locus allele combinations resulted in phenotypes that deviated significantly from the expectations, had the loci acted in an additive manner. Critically, this thesis demonstrates that non-additive genetic mechanisms often need to be considered in order to fully understand the genetics of complex traits.

Widespread unexplained variations in clinical practices and patient outcomes, together with rapidly growing availability of data, suggest major opportunities for improving the quality of medical care. One way that healthcare practitioners try to do that is by participating in organized healthcare quality improvement collaboratives (QICs). In QICs, teams of practitioners from different hospitals exchange information on clinical practices, with the aim of improving health outcomes at their own institutions. However, what works in one hospital may not work in others with different local contexts, due to non-linear interactions among various demographics, treatments, and practices. I.e., the clinical landscape is a complex socio-technical system that is difficult to search. In this dissertation we develop methods for analysis and modeling of complex systems, and apply them to the problem of healthcare improvement. Searching clinical landscapes is a multi-objective dynamic problem, as hospitals simultaneously optimize for multiple patient outcomes. We first discuss a general method we developed for finding which changes in features may be associated with various changes in outcomes at different points in time with different delays in affect. This method correctly inferred interactions on synthetic data, however the complexity and incompleteness of the real hospital dataset available to us limited the usefulness of this approach. We then discuss an agent-based model (ABM) of QICs to show that teams comprising individuals from similar institutions outperform those from more diverse institutions, under nearly all conditions, and that this advantage increases with the complexity of the landscape and the level of noise in assessing performance. We present data from a network of real hospitals that provides encouraging evidence of a high degree of similarity in clinical practices among hospitals working together in QIC teams. Based on model outcomes, we propose a secure virtual collaboration system that would allow hospitals to efficiently identify potentially better practices in use at other institutions similar to theirs, without any institutions having to sacrifice the privacy of their own data. To model the search for quality improvement in clinical fitness landscapes, we need benchmark landscapes with tunable feature interactions. NK landscapes have been the classic benchmarks for modeling landscapes with epistatic interactions, but the ruggedness is only tunable in discrete jumps. Walsh polynomials are more finely tunable than NK landscapes, but are only defined on binary alphabets and, in general, have unknown global maximum and minimum. We define a different subset of interaction models that we dub as NM landscapes. NM landscapes are shown to have smoothly tunable ruggedness and difficulty and known location and value of global maxima. With additional constraints, we can also determine the location and value of the global minima. The proposed NM landscapes can be used with alphabets of any arity, from binary to real-valued, without changing the complexity of the landscape. NM landscapes are thus useful models for simulating clinical landscapes with binary or real decision variables and varying number of interactions. NM landscapes permit proper normalization of fitnesses so that search results can be fairly averaged over different random landscapes with the same parameters, and fairly compared between landscapes with different parameters. In future work we plan to use NM landscapes as benchmarks for testing various algorithms that can discover epistatic interactions in real world datasets.

Mémoire numérisé par la Division de la gestion de documents et des archives de l'Université de Montréal.

Considerado um dos aspectos mais complexos da genética quantitativa, a epistasia tem sido ignorada pelos melhoristas nos estudos de herança dos caracteres, principalmente os da herança da resistência de plantas a insetos, que são de difícil obtenção. No milho, a principal praga de grãos é o Sitophilus zeamais (Coleoptera: Curculionidae), devido a sua capacidade de atacar grãos tanto no campo quanto em silos. Contudo, as estimativas dos componentes aditivo e de dominância envolvidos na herança dessa resistência podem estar viesadas pela presença do efeito da epistasia. Utilizando o delineamento triple testcross, os objetivos deste trabalho foram: (i) verificar a presença da epistasia para os caracteres relacionados à resistência do milho ao S. zeamais; (ii) estimar os efeitos epistáticos em cada planta F2; e (iii) estimar o efeito da interação epistasia x ambientes para estes caracteres. As 300 progênies de retrocruzamentos utilizadas nesse estudo foram avaliadas em dois ambientes no município de Piracicaba/SP, em delineamento alfa-látice 15 x 20, no esquema fatorial com duas repetições por ambiente. Os caracteres avaliados foram: número de insetos mortos (NM); número de insetos emergidos (EM); tempo médio de desenvolvimento dos insetos (TM); índice de suscetibilidade (IS) e perda percentual de massa seca dos grãos (PE). No ambiente E. E. Anhumas a presença da epistasia foi detectada para todos os caracteres; porém, no ambiente Caterpillar o efeito da epistasia não foi detectado para nenhum dos caracteres. Na análise conjunta, os efeitos epistáticos foram detectados para os caracteres NM, EM, IS e PE. O efeito da epistasia do tipo aditivo x dominante e/ou dominante x dominante foi mais importante para todos os caracteres que a epistasia do tipo aditiva x aditiva. A interação da epistasia com os ambientes foi significativa apenas para os caracteres TM e PE. Identificaram-se efeitos epistáticos bidirecionais significativos em plantas F2 para todos os caracteres no ambiente E. E. Anhumas e para os caracteres NM, EM, IS e PE na análise conjunta. O grande número de plantas F2 que apresentaram epistasia para mais de um caráter simultaneamente sugere a presença de epistasia pleiotrópica. As estimativas da variância aditiva e da interação aditiva com ambientes foram significativamente maiores que as das variâncias de dominância para a maioria dos caracteres. As magnitudes das estimativas dos coeficientes de herdabilidade para todos os caracteres variaram de baixas a medianas. A alta correlação genética entre os caracteres EM e PE sugere que o caráter PE, que é de difícil avaliação, pode ser selecionado indiretamente através do caráter EM que é de fácil avaliação para programas de melhoramento visando resistência ao S. zeamais. Os resultados obtidos na análise conjunta sugerem que, na população estudada, a epistasia constitui um componente importante da variância genética, de forma que as estimativas da variância aditiva, de dominância, graus médios de dominância e coeficientes de herdabilidade estão viesadas. / Considered one of the most complex components in quantitative genetics, the epistasis has been ignored by plant breeders, especially in inheritance studies of plant resistance, because the traits are laborious to evaluate. Sitophilus zeamais (Coleoptera: Curculionidae) is an important grain pest, due to its ability to attack in both field and silo conditions. However, the estimation of additive and dominance components in the inheritance of resistance to this pest may be biased due to epistatic effects. With the use of the triple test cross design, this research was aimed to: (i) verify the role of epistasis in the inheritance of maize resistance to S. zeamais, (ii) estimate the epistatic effects in the resistance traits of each F2 plant; (iii) estimate the epistasis x environment interaction. The 300 backcross progenies of this study were evaluated in two environments at Piracicaba, Brazil, in 2008/2009 growing season, using an alpha-lattice 15 x 20 design in a factorial arrangement with two replications per environment. The recorded traits were: number of dead weevils (NDW); number of emerged weevils (EW); mean development period (DP); index of susceptibility (IS) and the percentage of dry grain weight loss (DGWL). The epistatic effects were detected in Anhumas Experimental Station (AES) environment for all traits although they were absent in Caterpillar Experimental Station environment. In the combined analysis epistasis was detected for NDW, EW, IS and DGWL traits. The additive x dominance and/or dominance x dominance epistasis were more important than the additive x additive epistasis for all traits. The epistasis x environment interaction was significant only for traits DP and DGWL. Significant epistatic effects, which were not unidirectional, were detected in F2 plants for all traits in AES and for the traits NDW, EW, IS and DGWL in the combined analysis. Several F2 plants presented epistasis for more than one trait simultaneously suggesting the presence of pleiotropic epistasis. Estimates of additive, dominance and the additive by environment interaction variances were significant for all traits. Estimates of additive and additive by environment interaction variances were significantly higher than those of dominance variance for most of the traits. The magnitudes of the heritability coefficients estimates ranged from low to intermediate. The genetic correlation between EW and DGWL traits suggests that DGWL, which is difficult to evaluate, can be indirectly selected by the EW trait, which is easier to evaluate, in a breeding program. The results from the combined analysis suggests that, in the population studied, the epistasis is an important component of the genetic variance; therefore, estimates of additive, dominance variance, the average levels of dominance and heritability coefficients are biased.

Este estudo teve por objetivo mapear QTLs para características de desempenho e de carcaça em Gallus gallus . Foram estudadas 350 aves F2 oriundas de um cruzamento, na primeira geração, de machos de corte da linhagem TT com fêmeas de postura da linhagem CC. O peso vivo com 1, 35 e 42 dias de idade; o ganho de peso, o consumo de ração e a conversão alimentar de 35 a 41 dias de idade; os pesos dos pulmões, fígado, coração, moela, peito, coxas (peso de coxas e sobre-coxas), carcaça (sem vísceras, pés e cabeça), carcaça residual (peso da carcaça sem peito, asas e coxas), asas, cabeça, pés e gordura abdominal; o comprimento do intestino e o percentual de hematócrito, foram os fenótipos analisados. Foram utilizados 79 marcadores microssatélites, os quais cobriram 1510,7 cM dos cromossomos 1, 2, 3, 4, 5, 8, 11 e 13. Primeiramente, foram realizadas análises isoladas de cada fenótipo original e de variáveis canônicas obtidas por análise de componentes principais dos fenótipos. O teste razão de verossimilhanças (LRT) entre um modelo incompleto (apenas com efeitos fixos de sexo, incubação e o efeito aleatório de valor genético infinitesimal) e um completo (todos os efeitos anteriores mais os efeitos de QTL) foi o procedimento utilizado nas análises, exceto para testar modelos com interações epistáticas, onde a metodologia de quadrados mínimos foi utilizada. Modelos com interação QTL x sexo também foram testados. Posteriormente, foram feitas análises de múltiplos fenótipos simultaneamente, onde foi possível testar a hipótese de QTL pleiotrópico x QTLs ligados, além dos testes descritos acima, com exceção de efeitos epistáticos. As análises descritivas e de componentes principais foram obtidas no SAS, enquanto o mapeamento de QTL foi realizado no programa QxPak, exceto para análise de efeitos epistáticos, em que um código em Fortran 90 foi empregado. O modelo univariado, sem interações, permitiu mapear oito QTLs altamente significativos (cinco no GGA1 para PV35, PV42, gordura abdominal, comprimento do intestino e peso da cabeça; dois QTLs no GGA2 para PV35 e PV42; e um QTL no GGA3 para gordura abdominal) seis significativos (dois no GGA1 para conversão alimentar e ganho de peso; dois no GGA3 para peso das asas e das coxas; um no GGA4 para peso da cabeça; e um no GGA8 para peso da moela), além de 13 ligações sugestivas para diversas características. Dez QTLs apresentaram interação com sexo, sendo cinco específicos para machos. O modelo com busca simultânea de dois QTLs mapeou seis QTLs anteriormente perdidos (cinco para PV35 e PV42; e um para peso da cabeça). Interações epistáticas foram observadas para PV35 e PV42 entre um QTL em 69 cM do GGA1 com QTLs em 333 cM do GGA1, 272 cM do GGA3 e 77 cM do GGA5. Dois QTLs e seis ligações sugestivas foram mapeados na análise de variáveis canônicas, os quais não haviam sido mapeados com as variáveis originais. Com o procedimento de múltiplas características foi possível mapear nove QTLs pleiotrópicos e o aumento de poder do teste foi evidenciado, principalmente, no GGA2. / This study aim to map QTL for performance and carcass traits in (Gallus gallus) . There were used 350 F2 chickens developed by crossing a broiler male line (TT) with a layer line (CC). The body weight with 1, 35 and 42 days of age, weight gain, feed intake and feed conversion from 35 to 41 days, weights of lung, liver, heart, gizzard, breast, drums and thighs, carcass (without giblets, feet and head), residual carcass (weight of carcass without breast, drums, thighs, and wings), wings, head, feet, and abdominal fat, intestine length and hematócrito value were the phenotypes analyzed. Seventy nine microssatellite markers were used, which covered 1510.7 cM of chromosomes 1, 2, 3, 4, 5, 8, 11, and 13. Firstly, QTL analysis was carried out for each original trait and for canonical variables, obtained from principal components analysis of the phenotypes. The likelihood ratio test (LRT) between a reduced model (only fixed effects of sex, hatch and random effect of infinitesimal genetic value) and a full model (all anterior effects and QTL effects) was applied to map QTL, but mean square approach was used for mapping QTL with epistatic effect. Besides, models with QTL by sex interaction were also tested. Finally, multi-trait analysis was used to test the hypothesis of pleiotropic x linkage QTLs, besides of the tests previously described, except models with epistatic effects. For descriptive and principal components analysis the SAS software was used. QTL mapping was carried out with QxPak software and a fortran 90 source code to test models with epistatic effect. The univariate model, without interactions, allowed to map eight highly significant QTLs (five in the GGA1, for PV35, PV42, abdominal fat, intestine length, and head weight; two QTLs in the GGA2, for PV35 and PV42; and one QTL in the GGA3 for abdominal fat), six significant QTLs (two in the GGA1 for feed conversion and weight gain; two in the GGA3 for wings and drums and thighs weights; one in the GGA4 for head weight; and one in the GGA8 for gizzard weight), besides 13 suggestive linkages for several traits. Ten QTLs interacted with sex, being five of them male specific QTLs. The model with simultaneous search for two QTLs was important to map six QTLs previously lost (five for body weight at 35 and 42 days; and one for head weight). Epistatic Interactions were observed for body weight among a QTL in 69 cM of GGA1 with QTLs in 333 cM of GGA1, 272 cM of GGA3 and 77 cM of GGA5. Two QTLs and six suggestive linkages were mapped with the analysis on canonical variables, which have not been mapped with the original variables. With the multi-trait approach nine pleiotropic QTLs were mapped and an increase in the test power was observed mainly in the GGA2 chromosome.

Les maladies chroniques, comme les maladies cardiovasculaires (MCV), la maladie d'Alzheimer (AD), la dépression et l'ostéoporose, sont les principales causes de mortalité dans le monde. L'identification de facteurs de risque communs à ces maladies pourrait contribuer à un vieillissement «sain» mieux surveillé en utilisant des stratégies personnalisées de prédiction des risques, de prévention précoce et de traitement adéquat, en tenant compte des comorbidités très souvent existantes. Dans cette thèse, 8 publications ont été développées. Dans un premier temps, j'ai résumé, dans un article de revue, les défis actuels et les opportunités de la pharmacogénomique des médicaments contre les maladies cardiovasculaires. J'ai participé à la formation d'un consortium international, le Consortium VEGF et j'ai participé à une étude qui a identifié des interactions épistasiques entre les polymorphismes qui régulent les niveaux de VEGF et la pression artérielle et les indices d'adiposité. J'ai également démontré qu’un marqueur génétique de VEGF, le rs4416670, était significativement associé à un risque accru de dépression. En outre, j'ai signalé deux interactions significatives entre les variantes liées au VEGF affectant la densité minérale osseuse du col fémoral chez les femmes ménopausées. J'ai également étudié deux marqueurs liés au métabolisme des lipides : l'apolipoprotéine E (APOE) et le «lipolysis-stimulated receptor» (LSR). J'ai trouvé que le variant LSR rs916147 peut interagir avec APOE d'une manière qui inverse l'effet protecteur de l'allèle ε2 de l'APOE sur les lipides sanguins, fournissant ainsi de nouvelles connaissances sur les mécanismes de l'hyperlipoprotéinémie de type III. Les interactions épistasiques entre ces deux gènes augmentent également le risque d’AD même en l'absence de l'allèle à risque, APOE ε4. Finalement, j'ai réalisé des études épigénetiques (EWAS) sur l'obésité centrale et les traits lipidiques chez des individus sains. Les résultats suggèrent qu'un CpG pourrait affecter le tour de taille à travers une voie de signalisation de l'insuline. En outre, deux CpGs ont été associées aux niveaux des triglycérides par des gènes liés aux maladies cardiaques génétiques (PRKAG2) et à l'inhibition de la signalisation Wnt / bêta-caténine impliquée dans le développement des MCV et d’AD (KREMEN2). En conclusion, dans cette thèse j’ai utilisé l'étude de l'épistasie et de l'épigénétique pour identifier des interrelations complexes entre VEGF, LSR, APOE et différentes maladies chroniques (MCV, AD, ostéoporose, dépression) proposant ainsi de nouveaux mécanismes et des dénominateurs communs de ces maladies qui devraient être utilisés comme biomarqueurs de médecine personnalisée / Chronic diseases, like cardiovascular diseases (CVD), Alzheimer’s disease (AD), depression and osteoporosis, are major causes of mortality in the world. Identification of common to those diseases risk factors could help for a better-monitored ‘healthy’ aging, by promotion of personalised strategies for risk prediction, early prevention and adequate treatment, all taking into account the very often existing comorbidities. In this thesis, 8 publications have been developed. Initially, in a review paper, I have summarised the current challenges and opportunities of pharmacogenomics of CVD medications. I have participated in the formation of an international consortium, the VEGF Consortium, and I have participated in a study that identified significant epistatic interactions between polymorphisms that regulate the levels of VEGF and their effects on blood pressure and adiposity indexes. I have also demonstrated that one genetic marker of VEGF, rs4416670, was significantly associated with an increased risk for depression. Furthermore, I have reported two significant interactions between VEGF-related variants affecting the femoral neck bone mineral density in post-menopausal women. I have focused also on two markers linked with lipids metabolism: the apolipoprotein E (APOE) and the lipolysis-stimulated receptor (LSR). I have found that the LSR variant rs916147 can interact with APOE in a way that reverses the protective effect of the ε2 allele of APOE on blood lipids, thus providing new insights in the mechanisms underlying type III hyperlipoproteinemia. Epistatic interactions between these two genes have also been shown to increase the risk of AD, even in the absence of the known risk allele APOE ε4. Finally, I have performed epigenome-wide association studies (EWAS) on central obesity and blood lipid traits in healthy individuals. The results suggest that one methylation probe could affect waist circumference through an insulin-signaling pathway. Furthermore, two methylations probes were associated with triglycerides levels through genes linked with genetic heart diseases (PRKAG2) and with inhibition of the Wnt/beta-catenin signaling that is involved in CVD and AD development (KREMEN2). In conclusion, this thesis used the study of epistasis and epigenetics and identified complex inter-relationships between VEGF, LSR, APOE and different chronic diseases (CVD, AD, osteoporosis, depression) and novel mechanisms that link disease development with DNA methylation, thus demonstrating their role as common denominators of diseases that can be used as valuable markers in personalised medicine

High-dimensional data, such as that relating to gene expression in microarray experiments, may contain substantial amount of useful information to be explored. However, the information, relevant variables and their joint interactions are usually diluted by noise due to a large number of non-informative variables. Consequently, variable selection plays a pivotal role for learning in high dimensional problems. Most of the traditional feature selection methods, such as Pearson's correlation between response and predictors, stepwise linear regressions and LASSO are among the popular linear methods. These methods are effective in identifying linear marginal effect but are limited in detecting non-linear or higher order interaction effects. It is well known that epistasis (gene - gene interactions) may play an important role in gene expression where unknown functional forms are difficult to identify. In this thesis, we propose a novel nonparametric measure to first screen and do feature selection based on information from nearest neighborhoods. The method is inspired by Lo and Zheng's earlier work (2002) on detecting interactions for discrete predictors. We apply a backward elimination algorithm based on this measure which leads to the identification of many in influential clusters of variables. Those identified groups of variables can capture both marginal and interactive effects. Second, each identified cluster has the potential to perform predictions and classifications more accurately. We also study procedures how to combine these groups of individual classifiers to form a final predictor. Through simulation and real data analysis, the proposed measure is capable of identifying important variable sets and patterns including higher-order interaction sets. The proposed procedure outperforms existing methods in three different microarray datasets. Moreover, the nonparametric measure is quite flexible and can be easily extended and applied to other areas of high-dimensional data and studies.

DNA double-strand breaks (DSBs) containing unligatable termini are potent cytotoxic lesions leading to growth arrest or cell death. The Artemis nuclease and tyrosyl-DNA phosphodiesterase (TDP1) are each capable of resolving protruding 3′-phosphoglycolate (PG) termini of DNA double-strand breaks (DSBs). Consequently, a knockout of Artemis and a knockout/knockdown of TDP1 rendered cells sensitive to the radiomimetic agent neocarzinostatin (NCS), which induces 3′-PG-terminated DSBs. Unexpectedly, however, a knockdown or knockout of TDP1 in Artemis-null cells did not confer any greater sensitivity than either deficiency alone, indicating a strict epistasis between TDP1 and Artemis. Moreover, a deficiency in Artemis, but not TDP1, resulted in a fraction of unrepaired DSBs, which were assessed as 53BP1 foci. Conversely, a deficiency in TDP1, but not Artemis, resulted in a dramatic increase in dicentric chromosomes following NCS treatment. An inhibitor of DNA-dependent protein kinase, a key regulator of the classical nonhomologous end joining (C-NHEJ) pathway sensitized cells to NCS but eliminated the sensitizing effects of both TDP1 and Artemis deficiencies. Moreover, Polynucleotide Kinase/ Phosphatase (PNKP) is known to process 3′-phosphates and 5′-hydroxyls during DSB repair. PNKP-deficiency sensitized both HCT116 and HeLa cells to 3′-phosphate ended DSBs formed upon radiation and radiomimetic drug treatment. The increased cytotoxicity in the absence of PNKP was synonymous with persistent, un-rejoined 3′-phosphate-ended DSBs. However, DNA-PK deficiency sensitized PNKP-/- cells to low doses of NCS suggesting that, in the absence of PNKP, alternative enzyme(s) can remove 3′-phosphates in a DNA-PK-dependent manner. These results suggest that TDP1 and Artemis perform different functions in the repair of terminally blocked DSBs by the C-NHEJ pathway, and that whereas an Artemis deficiency prevents end joining of some DSBs, a TDP1 deficiency tends to promote DSB mis-joining. In addition, loss of PNKP significantly sensitizes cells to 3′-phosphate-ended DSBs due to a defect in 3′-dephosphorylation.

Epistasis describes non-additive interactions which affect gene expression and phenotype. It can happen on multiple levels, including on a genomic level with interactions between genes or even chromosomes affecting global patterns of gene expression. It can also happen within a gene itself, with epistatic interactions between amino acids affecting gene expression and resultant phenotypes. I present three studies in two organisms to study this phenomenon on a global-genomic scale, and also on a local-genic scale.

Mémoire numérisé par la Division de la gestion de documents et des archives de l'Université de Montréal