Os gliomas estão entre os mais freqüentes e fatais tumores do sistema nervoso central. Apesar dos grandes avanços da Medicina nas áreas diagnósticas e terapêuticas, o tratamento desses tumores ainda é, na grande maioria dos casos, paliativo, sendo a extirpação cirúrgica do tumor o único recurso com potencial curativo e, ainda assim, apenas para os gliomas de baixo grau. Neste trabalho, foram analisados os perfis de expressão gênica diferencial entre linhagens de glioma de rato com diferentes graus de tumorigenicidade (linhagens ST1 e P7), visando identificar genes com potencial de aplicação na doença humana como marcadores moleculares e/ou novos alvos terapêuticos. Numa primeira abordagem, foi realizado um rastreamento de genes potencialmente diferencialmente expressos entre as linhagens ST1 e P7 de glioma de rato através da hibridização de macroarrays de cDNA comerciais. Um conjunto de três membranas comerciais foi hibridizado com alvos radioativos gerados a partir de mRNA obtido destas duas linhagens, totalizando aproximadamente 3.000 genes. Nestes experimentos, foram identificados aproximadamente 200 genes como sendo possivelmente diferencialmente expressos entre as linhagens ST1 e P7. Os níveis de expressão relativa de cerca de 40 destes genes foram analisados por Northern blot, sendo que seis deles foram confirmados como sendo mais expressos na linhagem ST1 enquanto que 4 foram confirmados como sendo mais expressos na linhagem P7. Num segundo momento, foi realizado um rastreamento de expressão gênica em larga escala, através do uso de microarrays de DNA contendo sondas que permitem a análise de expressão de aproximadamente 24.000 transcritos de rato. Esta análise levou à identificação de uma lista contendo aproximadamente 1.300 genes candidatos a diferencialmente expressos, que apresentaram razões de expressão relativa entre 2 e >3.000. Desta lista, -700 genes foram identificados como provavelmente mais expressos na linhagem ST1, e cerca de 500 genes com expressão maior na linhagem P7. A etapa de subseqüente de validação da expressão diferencial foi realizada através de PCR quantitativo em tempo real (Q-PCR). A expressão de 49 genes foi quantificada em quatro preparações independentes de RNA originárias das linhagens ST1 e P7, sendo que 27 destes genes foram identificados como possivelmente diferencialmente expressos através dos microarrays, 10 dos quais haviam sido previamente confirmados por Northern Blot e 12 genes diversos. Destes genes, 21 foram confirmados por Q-PCR como sendo diferencialmente expressos entre as linhagens ST1 e P7, além daqueles 10 genes cuja expressão diferencial havia sido anteriormente confirmada por Northern. Ao todo, oito genes foram confirmados como sendo mais expressos na linhagem ST1, e 23 genes o foram como sendo mais expressos na linhagem P7. Vários dos genes confirmados como sendo diferencialmente expressos na linhagem ST1 estão, direta ou indiretamente, relacionados a parada de crescimento e supressão de fenótipo tumoral. Em contrapartida, diversos dos genes confirmados como sendo diferencialmente expressos na linhagem P7 codificam para receptores de fatores de crescimento peptídicos e seus respectivos ligantes. Em particular, foi detectado maior expressão, na linhagem P7, de receptores e Iigantes relacionados ao fatores de crescimento derivados de plaquetas (PDGFs), fatores de crescimento para fibroblastos (FGFs) e, também, do fatores de crescimento da família de pleiotrofina/midquina (PTN/MDK), e seus respectivos receptores. Sabidamente, vários destes genes estão envolvidos na neoangiogênese e na fechamento de alças autócrinas/parácrinas de estímulo da proliferação tumoral, em particular de gliomas humanos. Além disso, estes achados foram coerentes com o maior potencial tumorigênico (-2,5x) apresentado pela linhagem P7, quando injetadas s.c. no dorso de animais imunocomprometidos (p<O,01). As linhagens ST1 e P7 foram originalmente isoladas como sendo sublinhagens derivadas da linhagem C6, cuja origem é donal. Assim, uma possível interpretação para as diferenças marcantes dos perfis de expressão detectadas entre as linhagens ST1 e P7, é de que estas diferenças sejam, mais provavelmente, a conseqüência de algumas poucas alterações moleculares em genes-chave, e não de alterações extensas do genoma celular, uma vez que as duas linhagens têm, a priorí, o mesmo \"background\" genético. Assim, um número limitado de aberrações genéticas adicionais, particulares de cada linhagem, seria suficiente para uma modificação substancial dos perfis de expressão gênica, se os genes alterados forem capazes de modular a expressão de vários outros genes. Na busca de indícios que fortalecessem esta hipótese, foi quantificado, em seis linhagens humanas de glioma, a expressão dos mesmos genes cuja expressão foi originalmente quantificada nas linhagens ST1 e P7. Em seguida, foram realizados testes de correlação em pares (Teste de Pearson) entre os níveis de expressão relativos destes genes para as seis linhagens, buscando identificar conjuntos de genes que apresentassem expressão coordenada, que, por sua vez, sugeririam a existência de possíveis relações regulatórias entre os mesmos. Foi possível observar expressão significativamente correlacionada de seis genes (CHD7, FGF1, PDGFRA, PTN, PTPRZ1 e SCG2), sugerindo uma possível co-regulação recíproca entre diferentes vias de fatores de crescimento e/ou um novo fator remodelador de cromatina (CHD7). Pararelamente, o gene CHD7 foi, por sua vez, identificado como um novo fator remodelador de cromatina putativo, cujo cDNA completo pode ser amplificado, com êxito, através de RT-PCR longo. A expressão destes mesmos seis genes foi quantificada, por Q-PCR, em 64 amostras clínicas de glioma e cérebro humano não-tumoral. Com exceção do gene SCG2, os demais cinco (CHD7, FGF1, PDGFRA, PTN e PTPRZ1) são significativamente mais expressos em todos os graus de glioma, quando comparados com cérebro humano não-tumoral (p<0,05). Foi possível verificar, inclusive, que os níveis de expressão de vários destes genes estão altamente correlacionados nessas amostras. Em particular, observou-se correlação com alta significância entre os genes CHD7xPDGFRA (p=4,7x10-17), FGF1 xPTN (p=1, 1x10-19), do receptor PTPRZ1 contra todos os demais genes (100-6<p<10-9) e, especificamente, contra seu Iigante PTN (p=1,6x10-14). Como os loci dos genes PTN e PTPRZ1 se encontram relativamente próximos (-15 Mb) numa região do cromossomo 7 humano, a qual frequentemente se encontra amplificada em gliomas humanos (7q), é possível que estes dois genes, codificantes para um fator de crescimento e seu respectivo receptor, façam parte de um novo amplicon em gliomas humanos. Além disso, vários genes com maior expressão na linhagem P7 (ALK, FGFR1, RNF130 e ROS01) estão, também, significativamente mais expressos em certos graus de gliomas humanos, quando comparados com tecido cerebral humano não-tumoral. De um modo geral, estes dados indicam que foi possível obter, a partir de linhagens de glioma de rato, incursões aplicáveis para um entendimento mais amplo da biologia que envolve a patogênese da doença humana. Um destes dados extrapoláveis entre diferentes modelos inter-espécies é que a co-expressão de múltiplas vias autócrinas de crescimento parece ser um achado comum tanto em modelos experimentais de glioma em rato quanto em linhagens e amostras clínicas de gliomas humanos, e que estas vias estão, potencialmente, interconectadas ao nível transcricional. Além disso, foi possível verificar que um novo gene codificante para um fator de remodelamento de cromatina (CHD7), apresenta níveis de expressão relativos muito aumentados em amostras clínicas de glioma, quando comparado com cérebro humano não-tumoral. Analisados em conjunto, os resultados obtidos aqui têm o potencial para conduzir ao desenvolvimento racional de novas estratégias terapêuticas e diagnósticas/prognósticas para gliomas humanos. A análise do papel funcional destes genes em glioma, e das vias moduladas por seus produtos, se constituirá de um passo fundamental para o estabelecimento destes genes como potenciais novos alvos terapêuticos e/ou marcadores moleculares. / Gliomas are among the most frequent and fatal tumors of the central nervous system. Despite the recent and important advances in the diagnostic and therapeutic fields, treatment of these tumors still is, in the vast majority of the cases, palliative. Surgical ressection of the tumor remains as the sole potentially curative resource, but only for the low grade gliomas. In the present work, differential gene expression profiles were analyzed between rat glioma cell lines differing in tumorigenic potential (ST1 and P7 cell lines), aimed at identifying genes with potential to become novel molecular markers and therapeutic targets for the human disease. As a first approach, a commercial macroarray-based screening was undertaken in order to identify differentially expressed genes between ST1 and P7 rat glioma cell lines. Three different sets of commercial membranes comprising 3,000 genes were hybridized against radiolabeled targets that were synthesized from mRNA extracted from both cell lines. As a result, near 2,000 genes were identified as being possibly differentially expressed between ST1 and P7 celllines. The relative expression levels of 40 genes from this list were analyzed by Northern blot, and six genes were successfully confirmed as being expressed at higher levels in the ST1 cell line, whereas four genes confirmed heightened expression in P7 cells. As a second approach, a large-scale, Affymetrix microarray-based screening was performed, which allowed measuring the relative expression levels of about 24,000 rat transcripts. This analysis led to the identification of 1,300 candidate differentially expressed genes, for which the differential expression ratios ranged from 2 up to >3,000. From these, -700 genes displayed preferential expression in the ST1 cell line, while around 500 genes were more expressed in P7 cells. The following step, namely, validation of the differential expression, was achieved through Real-Time PCR (Q-PCR). Expression levels of 49 genes were measured in four independent RNA preparations from ST1 and P7 cell lines. About 27 of these genes were identified as putative differentially expressed , 10 of which had previously been confirmed by Northern blot as differentially expressed and 12 additional genes were also included. From this list, 21 genes were confirmed by Q-PCR as being diferentially expressed between the ST1 and P7 cell lines, in addition to those 10 whose differential expression was confirmed by Northern Slol. In total, eight genes were confirmed as being differentially expressed in the ST1 cell line, and 23 genes were confirmed as being expressed at higher leveis in the P7 cells. Many of the genes confirmed as being differentially expressed genes in the ST1 cell line are, directly or indirectly, related to growth arrest and suppression of the malignant phenotype. On the other hand, several of the genes displaying elevated expression in the P7 cell line code for growth factor ligands and receptors related to the PDGFs (platelet-derived growth factors), FGFs (fibroblast growth factors) and PTN/MDK (pleiotrophin/midkine) growth factor families. Many of these genes are already known as being related to neoangiogenesis and to the establishment of autocrine/paracrine loops in human gliomas. In addition, these findings are in keeping with the significantly higher (p<O.01) tumorigenic potential displayed by the P7 cellline (-2,5x), when both cell lines are injected s.c. in the dorsal region of immunodeficient nude mice. Moreover, ST1 and P7 celllines were originally isolated as clonal celllines from the C6 cell line which, in turn, is also acionai cell line. Therefore, a possible interpretation for the remarkably different gene expression profiles between ST1 and P7 cell lines is that those differences are, most Iikely, a consequence of a few molecular defects in key/master genes, rather than extensive aberrations of the cellular genome, given that those celllines have, a priori, similar genetic backgrounds. Thus, a limited number of genetic aberrations, characterisitic of each cell line, would be sufficient for a substantial modification of the gene expression profiles, provided that the altered genes are able to modulate the expression of each other. In an attempt to investigate theis point, the relative expression levels of the same genes were analyzed by Q-PCR, and pairwise correlation tests (Pearson correlation tests) were performed using expression data from the six human glioma cell lines, aimed at identifying gene sets that displayed coordinate expression whichwould suggest the existence of putative regulatory loops among them. It was possible to identify a c1uster of six genes (CHD7, FGF1, PDGFRA, PTN, PTPRZ1, SCG2) that displays coregulated expression, thus suggesting a possible reciprocal co-regulation among distinct growth factor pathways and a putative chromatin remodeling factor (CHD7). Furthermore, the CHD7 gene was identified as a novel, putative chromatin remodeling factor, and its full-Iength cDNA could be successfully amplified by means of long RTPCR. The expression levels of the same genes was quantified, by Q-PCR, in 64 clinicai samples, comprising gliomas and non-tumoral human brain tissue samples. Except for the SCG2 gene, the remaining five genes (CHD7, FGF1, PDGFRA, PTN and PTPRZ1) were expressed at significantly higher levels in glioma samples of all grades, when compared to non-tumoral human brain tissue samples (p<O.05). Likewise, we were able to verify that expression levels for many of these genes are strictly correlated in those samples. A particularly high levei of significance was found for the correlation of expression levels between CHD7xPDGFRA (p=4.7x10-17) and also for the FGF1xPTN gene pair (p=1.1x10-19). A strikingly high level of significance (p=1.6x10-14) was also found for the expression levels between the PTPRZ1 receptor and its PTN ligand. Given that the PTN and PTPRZ1 loei are relatively close to each other (-15 Mb) on the long arm of human chromosome 7, which, in turn, is frequently amplified in human gliomas, the genomic region including these two genes may well constitute a novel amplicon in human gliomas. In addition, several other genes with higher expression in the P7 cell line (ALK, FGFR1, RNF130 and ROB01) are also expressed at significantly higher levels in certain glioma grades, when compared to non-tumoral human brain tissue samples. Overall, these data indicate that starting from the rat glioma cell lines model, it was possible to obtain, insights applicable to a better understanding of the biology underlying the pathogenesis of human gliomas. Thus, co-expression of multiple autocrine loops seems to be a commonplace in the rat experimental glioma models as well as in the human glioma cell lines and in clinicai samples, where these pathways are potentially interconnected at the transcriptional leveI. Furthermore, we were able to detect increased mRNA expression levels of a novel putative chromatin remodelling factor (CHD7) in human glioma tissue samples, when compared to non-tumoral human brain tissue samples. Taken together, the results obtained in this work may potentially lead to the rational development of novel therapeutic, diagnostic and prognostic strategies for human gliomas. Functional analysis of these genes in glioma, and the pathways modulated by their products, will be a fundamental step for the establishment of these genes as potentially novel therapeutic targets and/or molecular markers.

O iodeto reduz a expressão, cauda poli(A) e taxa de tradução do mRNA da NIS, a partir de 30 min de tratamento. O estudo atual objetivou caracterizar as bases moleculares envolvidas nesses efeitos inibitórios. Células PCCl3 foram tratadas com NaI (10-3M), por diferentes períodos de tempo, e avaliou-se: a meia-vida do mRNA de NIS; o papel das porções 3\'UTR/5\'UTR; o envolvimento de eventos transcricionais; a organização do citoesqueleto de actina; o conteúdo total, meia-vida, degradação, internalização e atividade de NIS; a ativação da via PI3K/Akt. Os resultados evidenciaram que o excesso de iodeto: reduz a meia-vida e interage com a porção 3\'UTR do mRNA de NIS; inibe a atividade do promotor de NIS; desorganiza o citoesqueleto de actina; reduz o conteúdo total, de membrana, a meia-vida e atividade de NIS; aumenta a internalização de NIS por clatrinas, e sua degradação por lissosomos; ativa a via PI3K/Akt. Conclui-se que o iodeto ativa diferentes mecanismos moleculares, transcricionais e pós-transcricionais, para promover o efeito inibitório sobre a expressão de NIS. / Iodide reduces NIS mRNA expression, poly(A) tail length and transcription rate, after 30 min of treatment. The present study aimed to caracterize the molecular basis involved in these inhibitory effects. PCCl3 cells were treated with NaI (10-3M) and assays were performed to evaluate: NIS transcript half-life; the role of NIS mRNA 3\'UTR/5\'UTR; the involvement of transcriptional events; the organization of actin cytoskeleton; the total content, half-life, degradation, internalization and activity of NIS; the activation of PI3K/Akt pathaway. The results indicatred that iodide excess: reduces NIS transcript half-life and interacts with its 3\'UTR portion; inhibits NIS promoter activity; disrupts the actin cytoskeleton; reduces NIS total and membrane content, NIS half-life and NIS activity; induces the internalization of NIS through clathrin and its degradation through lisosomes; activates the PI3K/Akt pathway. In conclusion, iodide activates different molecular mechanisms, transcriptional and post-transcriptional, to promoter the inhibitory effect on NIS expression.

As bactérias usam proteínas intracelulares de estocagem de ferro (ferritinas) que permitem acesso ao metal quando está em baixa quantidade. Neste trabalho, o papel das ferritina Bfr e Dps foi estudado através da análise de fenótipo de linhagens mutantes, e sua regulação foi estudada utilizando o gene repórter lacZ, em ensaios de atividade de <font face=\"Symbol\">b-galactosidase. Os resultados mostraram aumento de expressão de bfr em excesso de ferro, e que este gene é positivamente regulado por Fur. A expressão de dps teve aumento em carência de ferro e na fase estacionária. Ensaios de expressão nas linhagens mutantes <font face=\"Symbol\">DoxyR, <font face=\"Symbol\">DsRNA1 e <font face=\"Symbol\">DsRNA2, mostraram diferenças nos níveis de expressão em comparação aos atingidos na linhagem selvagem. O fator sigma ECF SigJ é importante para níveis máximos de expressão de bfr e o fator SigT para máxima expressão de dps. Foram deletadas as regiões codificadoras de bfr e dps separadamente, ou ambas. Os resultados mostraram que o mutante <font face=\"Symbol\">Ddps e o mutante duplo apresentaram maior sensibilidade a H2O2, tendo a linhagem <font face=\"Symbol\">Dbfr padrão similar à linhagem NA1000. O ensaio de quantificação de ferro mostrou que as ferritinas têm papel equivalente na estocagem de ferro intracelular. / Bacteria utilize intracellular iron storage proteins (ferritins) that allow access to the metal when it is present in low amount. In this study the role of ferritins Bfr and Dps was studied by analyzing the phenotype of mutant strains, and their regulation was studied using the lacZ reporter gene, in <font face=\"Symbol\">b-galactosidase activity assays. The results showed increased expression of bfr in iron excess, and that this gene is positively regulated by Fur. dps expression was increased under iron deficiency and in the stationary phase. Expression assays in the mutant strains <font face=\"Symbol\">DoxyR, <font face=\"Symbol\">DsRNA1 and <font face=\"Symbol\">DsRNA2 showed differences in expression levels compared to wild type. ECF sigma factor SigJ is important for maximum levels of bfr expression and SigT for maximal dps expression. The coding regions of bfr, dps or both have been deleted. The results showed that the <font face=\"Symbol\">Ddps and double mutant strains showed increased susceptibility to H2O2, and <font face=\"Symbol\">Dbfr showed a similar pattern to the NA1000 strain. The iron concentration determination showed that both ferritins play equivalent roles in intracellular iron storage.

Previously studies used correlation of DNase I hypersensitivity sites sequencing (DNase-seq) experiments to predict interactions between enhancers and its target promoter gene. We investigate the correlation methods Pearson’s correlation and Mutual Information, using DNase-seq data for 100 cell-types in regions on chromosome one. To assess the performances, we compared our results of correlation scores to Hi-C data from Jin et al. 2013. We showed that the performances are low when comparing it to the Hi-C data, and there is a need of improved correlation metrics. We also demonstrate that the use of Hi-C data as a gold standard is limited, because of its low resolution, and we suggest using another gold standard in further studies.

No description available.

This project is an investigation of whether analysing subsets of time series gene expression data can give additional information about putatively co-regulated genes, compared to only using the whole time series. The original gene expression data set was partitioned into subsets and similarity was computed for both the whole timed series and subsets. Pearson correlation was used as similarity measure between gene expression profiles. The results indicate that analysing co-expression in subsets of gene expression data derives true-positive connections, with respect to co-regulation, that are not detected by only using the whole time series data. Unfortunately, with the actual data set, chosen similarity measure and partitioning of the data, randomly generated connections have the same amount of true-positives as the ones derived by the applied analysis. However, it is worth to continue further analysis of the subsets of gene expression data, which is based on the multi-factorial nature of gene regulation. E.g. other similarity measures, data sets and ways of partitioning the data set should be tried.

Recent methods to infer genetic networks are based on identifying gene interactions by similarities in expression profiles. These methods are founded on the assumption that interacting genes share higher similarities in their expression profiles than non-interacting genes. In this dissertation this assumption is validated when using mutual information as a similarity measure. Three algorithms that calculate mutual information between expression data are developed: 1) a basic approach implemented with the histogram technique; 2) an extension of the basic approach that takes into consideration time delay between expression profiles; 3) an extension of the basic approach that takes into consideration that genes are regulated in a complex manner by multiple genes. In our experiments we compare the mutual information distributions for profiles of interacting and non-interacting genes. The results show that interacting genes do not share higher mutual information in their expression profiles than non-interacting genes, thus contradicting the basic assumption that similarity measures need to fulfil. This indicates that mutual information is not appropriate as similarity measure, which contradicts earlier proposals.

Hepatocellular carcinoma (HCC) is the most common form of primary liver cancer, ranking the sixth most common cancer and third most common cause of cancer mortality worldwide. Alpha-fetoprotein (AFP) is a plasma protein that is highly expressed in the fetal liver and shut off after birth. AFP expression is elevated in regenerating adult liver and HCC and has been used extensively as a diagnostic marker of liver cancer. We have been studying mouse liver gene regulation to better understand mechanisms by which changes in gene expression contribute to liver development, homeostasis and disease. Zinc Fingers and Homeoboxes 2 (Zhx2) has been identified as a repressor of AFP, but the mechanism of this regulation remains unknown. Interestingly, all targets of Zhx2 that have been identified to date, including H19, Glypican 3, Elovl3 and Cytochrome P450 (CYP) genes, are also known to be misregulated in HCC. Thus, a better understanding of the mechanism by which these genes are regulated by Zhx2 will likely lead to new insights into gene regulation during HCC progression. Antisense transcripts belong to a diverse class of long noncoding RNA molecules > 200 nucleotides in length that often structurally resemble mRNAs, but do not encode proteins. While studying AFP mRNA regulation by Zhx2 in the mouse, our lab identified novel antisense AFP (asAFP) RNA transcripts that partially overlap the 3’ half of the mouse AFP gene. ENCODE tracks of ChIP-seq data for histone modifications in mouse liver show that the genomic region around the 5’ end of asAFP RNA has peaks for marks associated with promoters and enhancers. To better understand asAFP regulation, I identified the asAFP RNA 5’ end and the promoter elements that drive transcription. asAFP RNAs are ~5kb alternatively spliced, mainly cytoplasmic transcripts containing 2-4 exons. These transcripts were also detected in adult mouse liver RNA-seq data. asAFP is likely a noncoding RNA because it contains several small open reading frames that are 98 aa or smaller with no known functional domains or homology to known proteins. There is no evidence for similar transcripts in human liver. The abundance of asAFP RNA inversely correlates with AFP mRNA levels during postnatal liver development. Normally, asAFP RNA levels are high and AFP mRNA levels are low in the adult mouse liver. However, in the absence of Zhx2, AFP mRNA levels are higher and asAFP RNA levels are reduced, suggesting asAFP may be involved in the developmental regulation of AFP. Antisense transcripts function through a variety of mechanisms to positively or negatively regulate the expression of target genes. To explore the role of asAFP RNA in AFP gene regulation, I expressed segments of asAFP RNA in a mouse liver cell line and measured endogenous AFP mRNA levels. My data revealed that all segments of asAFP repressed endogenous AFP mRNA in trans. To determine the mechanism by which asAFP RNA regulates AFP, I expressed asAFP segments that overlapped only with exons or introns of AFP. The asAFP segments that overlap with the exons showed greater repression of endogenous AFP mRNA levels than those overlapping with intronic sequences. Additionally, I considered whether asAFP RNA repression of AFP mRNA may involve RNA editing by Adenosine deaminase acting on RNA (ADAR). ADARs convert adenosine to inosine in double-stranded RNAs that results in RNA degradation. My data indicate that AFP and asAFP dsRNA is not extensively edited, suggesting ADAR mediated decay is not involved in the regulation of AFP mRNA expression. However, further studies are required to determine the mechanism of cytoplasmic AFP mRNA degradation. Together, my data characterizes the transcriptional regulation of novel mouse asAFP transcripts and provides a model system to investigate how these transcripts regulate AFP mRNA through RNA-RNA interaction.

The human genome contains tens of thousands of gene loci which code for an even greater number of protein and RNA products. The highly complex temporal and spatial expression of these genes makes possible all the biological processes of life. Altered gene expression by mutation or deregulation is fundamental for the development of many human diseases. The ultimate aim of this thesis was to identify gene expression changes relevant to cancer. The advent of genome-wide expression profiling techniques, such as microarrays, has provided powerful new tools to identify such changes and researchers are now faced with an explosion of gene expression data. Processing, comparing and integrating these data present major challenges. I approached these challenges by developing and assessing novel methods for cross-platform analysis of expression data, scalable subspace clustering, and curation of experimental gene regulation data from the published literature. I found that combining results from different expression platforms increases reliability of coexpression predictions. However, I also observed that global correlation between platforms was generally low, and few gene pairs reached reasonable thresholds for high-confidence coexpression. Therefore, I developed a novel subspace clustering algorithm, able to identify coexpressed genes in experimental subsets of very large gene expression datasets. Biological assessment against several metrics indicates that this algorithm performs well. I also developed a novel meta-analysis method to identify consistently reported genes from differential expression studies when raw data are unavailable. This method was applied to thyroid cancer, producing a ranked list of significantly over-represented genes. Tissue microarray analysis of some of these candidates and others identified a number of promising biomarkers for diagnostic and prognostic classification of thyroid cancer. Finally, I present ORegAnno (www.oreganno.org), a resource for the community-driven curation of experimentally verified regulatory sequences. This resource has proven a great success with ~30,000 sequences entered from over 900 publications by ~50 contributing users. These data, methods and resources contribute to our overall understanding of gene regulation, gene expression, and the changes that occur in cancer. Such an understanding should help identify new cancer mechanisms, potential treatment targets, and have significant diagnostic and prognostic implications. / Medicine, Faculty of / Medical Genetics, Department of / Graduate

No description available.