Single nucleotide polymorphisms (SNPs) are DNA sequence variations, occurring when a single nucleotide –A, T, C or G – is altered. Arguably, SNPs account for more than 90% of human genetic variation. Dr. Tebbutt's laboratory has developed a highly redundant SNP genotyping assay consisting of multiple probes with signals from multiple channels for a single SNP, based on arrayed primer extension (APEX). The strength of this platform is its unique redundancy having multiple probes for a single SNP. Using this microarray platform, we have developed fully-automated genotype calling algorithms based on linear models for individual probe signals and using dynamic variable selection at the prediction level. The algorithms combine separate analyses based on the multiple probe sets to give a final confidence score for each candidate genotypes. Our proposed classification model achieved an accuracy level of >99.4% with 100% call rate for the SNP genotype data which is comparable with existing genotyping technologies. We discussed the appropriateness of the proposed model related to other existing high-throughput genotype calling algorithms. In this thesis we have explored three new ideas for classification with high dimensional data: (1) ensembles of various sets of predictors with built-in dynamic property; (2) robust classification at the prediction level; and (3) a proper confidence measure for dealing with failed predictor(s). We found that a mixture model for classification provides robustness against outlying values of the explanatory variables. Furthermore, the algorithm chooses among different sets of explanatory variables in a dynamic way, prediction by prediction. We analyzed several data sets, including real and simulated samples to illustrate these features. Our model-based genotype calling algorithm captures the redundancy in the system considering all the underlying probe features of a particular SNP, automatically down-weighting any ‘bad data’ corresponding to image artifacts on the microarray slide or failure of a specific chemistry. Though motivated by this genotyping application, the proposed methodology would apply to other classification problems where the explanatory variables fall naturally into groups or outliers in the explanatory variables require variable selection at the prediction stage for robustness.

Clustering microarray data is a helpful way of identifying genes which are biologically related. Unfortunately, when attempting to cluster microarray data, certain issues must be considered including: the uncertainty in the number of true clusters; the expression of a given gene is often a ected by the expression of other genes; and microarray data is usually high dimensional. This thesis outlines a Bayesian in nite Gaussian mixture model which addresses the issues outlined above by: not requiring the researcher to specify the number of clusters expected, applying a non-diagonal covariance structure, and using mixtures of factor analyzers and extensions thereof to structure the covariance matrix such that it is based on a few latent variables. This approach will be illustrated on real and simulated data.

In birds, estrogen is a strong stimulator of gene programs that regulate the formation of very low density lipoproteins (VLDL). Apolipoprotein-B (ApoB) is an integral part of very low density lipoproteins. In mammals, the rate of ApoB synthesis is controlled by post-translational means. In contrast, estrogen treated birds show changes in ApoB transcript level; in a natural setting, the bird?s metabolism and transcription are in great flux due to yolk formation. Besides the ApoB gene, the entire complement of genes that is necessary to form a VLDL is not known. To determine the genes that play a role in the formation of VLDL 7-10d old chicks were injected with estrogen at several time points over a 24hr period. Following exsanguinations by cardiac puncture, livers were removed and RNA was extracted. The RNA was quantified and hybridized to microarrays using a dual-dye system. Slides were scanned and analyzed, and features were extracted. To qualify microarray results, quantitative real time PCR (q-RTPCR) was done on a selection of genes. Previous studies had shown that approximately 200 genes are upregulated by the treatment of hormone naive chickens with estrogen. As a result of our liver transcriptional profiling, we identified 1,528 genes at 1.5hrs, 1,931 genes at 3hrs, 2,398 genes at 6hrs, 2,356 at 12hrs, and 1,713 genes at 24hrs following estrogen exposure. We determined that these regulated genes include those responsible for the transcription of RNA used to create the gene products that serve as components of VLDL itself or that act in VLDL assembly. These include genes encoding structural proteins, like ApoB, and genes encoding assembly-related proteins. Of the differentially expressed genes as compared to time 0, there were approximately 30% which were unannotated with regards to function limiting conclusions. We hope to determine the function of these genes and to annotate them based on this information.

Cancer is a devastating disease for human society with thousands of deaths and estimated new cases every year around the globe. Intensive research efforts on understanding the disease progression and determining effective diagnostics and therapeutics have been employed for over one hundred years. Throughout this time, and in particular during the last two decades, computational-based methods have gained increasing importance in cancer biology research by providing significant advantages in the analysis and interpretation of high-throughput data at the molecular and genomic levels. More specifically, after completion of the Human Genome Project in 2003, and with the Cancer Human Genome Project underway, high-throughput biological assays (e.g., microarray chips, next generation sequencing machines) have supplied researchers thousands of measurements per experimental sample. The massive amount of related data has oftentimes been challenging to interpret and translate, particularly in cancer biology and therapeutics. This thesis reports the results of three independent studies in which high-throughput gene expression is computationally analyzed to address longstanding issues in cancer biology. Two of the studies utilize data from ovarian cancer patients while the third involves data collected from pancreatic cancer patients. In Chapter 1, I address the importance of personalized profiling in pancreatic cancer ; in Chapter 2 the role of cancer stroma in the progression of ovarian cancer and in Chapter 3 evidence for the role of epithelial-to-mesenchymal transition (EMT) in ovarian cancer metastasis. More specifically, Chapter 1 emphasizes the power of personalized molecular profiling in unmasking unique gene expression signatures that correspond to each individual patient. These individual expression patterns (individual profiling), which may be overlooked by the traditional methods of gene signatures enriched in groups of afflicted individuals (group profiling), can provide valuable information for more successful targeted therapies. In order to address this issue in pancreatic cancer, comparisons of the most significantly differentially expressed genes and functional pathways were performed between cancer and control patient samples as determined by group vs. personalized analyses. There was little to no overlap between genes/pathways identified by group analyses relative to those identified by personalized analyses. These results indicated that personalized and not group molecular profiling is the most appropriate approach for the identification of putative candidates for targeted gene therapy of pancreatic and perhaps other cancers with heterogeneous molecular etiology. Chapter 2, also with strong implications on personalized molecular profiling, unveils the functional variability of the tumor microenvironment among ovarian cancer patients. The purpose of this study was to investigate the process of microenvironmental stroma activation in human ovarian cancer by molecular analysis of matched sets of cancer and surrounding stroma tissues from individual patients. Expression patterns of genes encoding signaling molecules and compatible receptors in the cancer stroma and cancer epithelia samples indicated the existence of two sub-groups of cancer stroma with different propensities to support tumor growth. These results demonstrated that functionally significant variability exists among ovarian cancer patients in the ability of the microenvironment to modulate cancer development. Chapter 3 aims to uncover the molecular mechanisms that underlie the metastatic process with the hope that such knowledge may lead to more effective therapeutic treatments. For this purpose, pathological and molecular analyses were conducted in 14 matched sets of primary and metastatic samples from late staged ovarian cancer patients. Pathological examination revealed no morphological differences between any of the primary and metastatic samples. In contrast, gene expression analyses identified two distinct groups of patient samples. One group displayed essentially identical expression patterns to primary samples isolated from the same patients. The second group displayed expression patterns significantly different from primary samples isolated from the same patients. Predominant among the differentially expressed genes characterizing this second class of metastatic samples were genes previously associated with epithelial-to-mesenchymal transtion (EMT). These results supported a role of EMT in at least some ovarian cancer metastases and demonstrated that indistinguishable morphologies between primary and metastatic cancer samples is not sufficient evidence to negate the role of EMT in the metastatic process. / The data related to the ovarian cancer work discussed in this dissertation are available at: http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE38666

Single nucleotide polymorphisms (SNPs) are DNA sequence variations, occurring when a single nucleotide –A, T, C or G – is altered. Arguably, SNPs account for more than 90% of human genetic variation. Dr. Tebbutt's laboratory has developed a highly redundant SNP genotyping assay consisting of multiple probes with signals from multiple channels for a single SNP, based on arrayed primer extension (APEX). The strength of this platform is its unique redundancy having multiple probes for a single SNP. Using this microarray platform, we have developed fully-automated genotype calling algorithms based on linear models for individual probe signals and using dynamic variable selection at the prediction level. The algorithms combine separate analyses based on the multiple probe sets to give a final confidence score for each candidate genotypes. Our proposed classification model achieved an accuracy level of >99.4% with 100% call rate for the SNP genotype data which is comparable with existing genotyping technologies. We discussed the appropriateness of the proposed model related to other existing high-throughput genotype calling algorithms. In this thesis we have explored three new ideas for classification with high dimensional data: (1) ensembles of various sets of predictors with built-in dynamic property; (2) robust classification at the prediction level; and (3) a proper confidence measure for dealing with failed predictor(s). We found that a mixture model for classification provides robustness against outlying values of the explanatory variables. Furthermore, the algorithm chooses among different sets of explanatory variables in a dynamic way, prediction by prediction. We analyzed several data sets, including real and simulated samples to illustrate these features. Our model-based genotype calling algorithm captures the redundancy in the system considering all the underlying probe features of a particular SNP, automatically down-weighting any ‘bad data’ corresponding to image artifacts on the microarray slide or failure of a specific chemistry. Though motivated by this genotyping application, the proposed methodology would apply to other classification problems where the explanatory variables fall naturally into groups or outliers in the explanatory variables require variable selection at the prediction stage for robustness.

Although exposed to similar environmental stimuli, not all humans develop asthma. Similarly, mouse strains vary in the degree of pathophysiology seen following induction of experimental asthma. A model involving immunization and aerosol challenge with ovalbumin (OVA) was used to investigate factors that may confer strain-dependent resistance or susceptibility to pathology. BALB/c and C57BL/6 mice developed many features of human asthma including inflammation, mucus production and airway obstruction. In contrast, CBA/Ca mice were relatively resistant to development of disease. This was despite the presence of a robust systemic allergic response, as indicated by high levels of OVA-specific and total immunoglobulin and increases in circulating eosinophils comparable to those in BALB/c and C57BL/6 mice. In interleukin (IL)-5 transgenic (Tg) mice the strain specific susceptibility to lung mucus production and airway obstruction was maintained and pathology was greatly accentuated in C57BL/6 and BALB/c but not in CBA/Ca mice. Eosinophils recovered by bronchoalveolar lavage (BAL) from wt and IL-5 Tg CBA/Ca mice lost viability faster than BAL eosinophils from the other two strains and this phenomenon was lung-specific. This may result in less eosinophil accumulation in the lungs of CBA/Ca mice and resistance to asthma-like pathology. Fl hybrids of CBA/Ca mice crossed with either BALB/c or C57BL/6 mice had BAL leukocyte, eosinophil lifespan and cell-free protein profiles similar to those of the respective disease-susceptible parental strains. It is likely that eosinophil apoptosis was not mediated through the extrinsic or receptor mediated pathway. Bcl-2 and Bcl-xL, which both inhibit the intrinsic pathway of apoptosis were highest in BAL eosinophils from the BALB/c strain and this correlated with relatively high IL-5 levels in the lungs. Survivin inhibits apoptosis and expression was significantly higher in BALB/c and C57BL/6 BAL eosinophils than in cells from CBA/Ca mice. This suggests a possible mechanism whereby eosinophils from the asthma-susceptible C57BL/6 and BALB/c mice are more resistant to apoptosis and may account, in part, for the more extensive pathology in these strains. Using global gene expression analysis we identified groups of genes that were differentially regulated in the lungs of mice that are susceptible or resistant to development of asthma-like pathology. 242, 145 and 42 genes were differentially regulated in the lungs of the C57BL/6, BALB/C and CBA/Ca strains respectively. In C57BL/6 mice, transcripts were significantly enriched for adhesion molecules and we postulate that heightened expression of L-selectin, CD 18, PGSL-1 and LPAM-l on lung eosinophils is responsible for robust recruitment and therefore accumulation of these cells in C57BL/6 mice. 64 genes were differentially regulated only in the asthma-susceptible strains, several of which have not previously been associated witb asthma. The late expression of Chi313, Retnla and Mmp12 correlated with increased expression of IL-10 in the lungs and we hypothesise that this cytokine may be produced by alternatively activated macrophages as part of the resolution of disease. This study identifies several novel genes and mechanisms associated with the modulation of airway inflammation and pathology. The identification of factors that control allergic inflammation may provide novel therapeutic targets for disease intervention. / http://proxy.library.adelaide.edu.au/login?url= http://library.adelaide.edu.au/cgi-bin/Pwebrecon.cgi?BBID=1366239 / Thesis (Ph.D.) -- University of Adelaide, School of Molecular and Biomedical Science, 2009

Philadelphia positive (Ph+) Acute Lymphoblastic Leukaemia (ALL) is characterised by the presence of the BCR-ABL fusion gene, which encodes a protein tyrosine kinase with aberrant activity. Imatinib, a chemical Bcr-Abl inhibitor, is rarely effective in Ph+ ALL patients as a single agent. In this study, insight into molecular and signalling changes occurring in Ph+ ALL during Imatinib therapy were investigated using cDNA microarrays. An optimal microarray assay was established to examine the gene expression changes in leukaemic cells from Ph+ ALL patients treated with Imatinib. Over 500 genes with ≥1.5-fold up- or down-regulation were identified. Based on gene ontology and novelty to Bcr-Abl signalling, six genes were selected and expression changes in five of these genes (PKCε, PINK1, SPRY2, ATF4 and PECAM1) confirmed by real time RT-PCR in Imatinib treated primary Ph+ ALL cells or the SUP-B15 cell line. The functional role of Protein Kinase C epsilon (PKCε) in response to Imatinib was further investigated using the Ph+ lymphoid and myeloid cell lines, SUP-B15 and K562. Detection of Imatinib-induced apoptosis by annexin V and PI staining demonstrated that SUP-B15 cells were less sensitive to Imatinib compared to K562 cells. PKCε mRNA was 50-fold higher in Ph+ ALL cells than Ph+ myeloid cells. In SUP-B15 cells, Imatinib upregulated PKCε mRNA but the protein was reduced by proteolytic cleavage. Inhibition of caspases showed that this cleaved product was not required for Imatinib induced-apoptosis. The treatment of SUP-B15 and primary Ph+ ALL cells with TAT-εV1-2 peptide, a specific inhibitor of PKCε, increased Imatinib-induced apoptosis. While the forced overexpression of PKCε in K562 cells reduced Imatinib-induced apoptosis. This increased expression of PKCε was associated with the increase of survival and anti-apoptotic proteins, Akt and Bcl-2. In summary, Gene expression profiling of Ph+ ALL cells during Imatinib therapy identified PKCε as an Imatinib responsive gene. A novel role of PKCε in Ph+ ALL response to imatinib is proposed. Experimental data presented in this thesis indicate that PKCε mediates pro-survival/anti-apoptosis signals in Ph+ ALL thereby reducing Imatinib-induced death. Thus, targeting PKCε during Imatinib therapy may be beneficial for the future treatment of Ph+ ALL.

Colorectal carcinoma arises through sequential genetic changes whereby an adenoma develops from normal colonic epithelium and then becomes a carcinoma. Critical to this process is two forms of mutually exclusive genomic instability ?? chromosomal instability (CIN) and microsatellite instability (MSI). The colorectal tumours that develop from each of these pathways have distinct pathological and molecular differences. Most MSI+ colorectal carcinomas are associated with the CpG island methylator phenotype (CIMP) - an epigenetic phenomenon where a specific and consistent group of genes are silenced through promoter methylation. However, over half of fall CIMP+ colorectal tumours are microsatellite stable (MSS). It is well known that the WNT/β-catenin signalling pathway is instrumental in the initiation and development of CIN type tumours but it is less clear whether this pathway has any significant involvement in MSI+ or methylated tumours. The role of the PI3K1AKT signalling pathway in the development of solid human tumours has only recently been established and the affects of abnormal PI3K/AKT signalling in sporadic colorectal carcinomas is yet to be fully elucidated. The objective of this thesis was to investigate the involvement of the WNT/β-catenin and PI3K/AKT signalling pathways in the CIN, MSI+ and methylated subgroups of sporadic colorectal carcinoma. To achieve this, the expression patterns of β-catenin, p-AKT and PTEN were identified by immunohistochemistry on sections from tissue microarrays consisting of cores from a large group of sporadic colorectal carcinomas. Each of these proteins is an integral part of the constitutive activation of WNT/β-catenin or PI3K/AKT signalling and their expression patterns were correlated with the clinical, pathological and molecular characteristics of the different subgroups of colorectal carcinoma. Increased nuclear β-catenin expression, an indicator of activated WNT signalling, is associated with MSS and the pathological features of CIN type tumours and inversely associated with the pathological and molecular features of MSI+ and CIMP+ tumours. In all forms of sporadic colorectal carcinoma, nuclear β-catenin expression was not an indicator of overall survival. PTEN was not associated with any particular subgroup of sporadic colorectal carcinoma, but decreased cytoplasmic expression was indicative of overall worse outcome, especially in MSS or CIN type tumours. While the identification of nuclear β-catenin in sporadic colorectal carcinomas is not a satisfactory prognostic marker, the immunohistochemical detection of absent PTEN expression may prove useful in identifying poor outcome in individuals with sporadic MSS colorectal carcinoma.

Although exposed to similar environmental stimuli, not all humans develop asthma. Similarly, mouse strains vary in the degree of pathophysiology seen following induction of experimental asthma. A model involving immunization and aerosol challenge with ovalbumin (OVA) was used to investigate factors that may confer strain-dependent resistance or susceptibility to pathology. BALB/c and C57BL/6 mice developed many features of human asthma including inflammation, mucus production and airway obstruction. In contrast, CBA/Ca mice were relatively resistant to development of disease. This was despite the presence of a robust systemic allergic response, as indicated by high levels of OVA-specific and total immunoglobulin and increases in circulating eosinophils comparable to those in BALB/c and C57BL/6 mice. In interleukin (IL)-5 transgenic (Tg) mice the strain specific susceptibility to lung mucus production and airway obstruction was maintained and pathology was greatly accentuated in C57BL/6 and BALB/c but not in CBA/Ca mice. Eosinophils recovered by bronchoalveolar lavage (BAL) from wt and IL-5 Tg CBA/Ca mice lost viability faster than BAL eosinophils from the other two strains and this phenomenon was lung-specific. This may result in less eosinophil accumulation in the lungs of CBA/Ca mice and resistance to asthma-like pathology. Fl hybrids of CBA/Ca mice crossed with either BALB/c or C57BL/6 mice had BAL leukocyte, eosinophil lifespan and cell-free protein profiles similar to those of the respective disease-susceptible parental strains. It is likely that eosinophil apoptosis was not mediated through the extrinsic or receptor mediated pathway. Bcl-2 and Bcl-xL, which both inhibit the intrinsic pathway of apoptosis were highest in BAL eosinophils from the BALB/c strain and this correlated with relatively high IL-5 levels in the lungs. Survivin inhibits apoptosis and expression was significantly higher in BALB/c and C57BL/6 BAL eosinophils than in cells from CBA/Ca mice. This suggests a possible mechanism whereby eosinophils from the asthma-susceptible C57BL/6 and BALB/c mice are more resistant to apoptosis and may account, in part, for the more extensive pathology in these strains. Using global gene expression analysis we identified groups of genes that were differentially regulated in the lungs of mice that are susceptible or resistant to development of asthma-like pathology. 242, 145 and 42 genes were differentially regulated in the lungs of the C57BL/6, BALB/C and CBA/Ca strains respectively. In C57BL/6 mice, transcripts were significantly enriched for adhesion molecules and we postulate that heightened expression of L-selectin, CD 18, PGSL-1 and LPAM-l on lung eosinophils is responsible for robust recruitment and therefore accumulation of these cells in C57BL/6 mice. 64 genes were differentially regulated only in the asthma-susceptible strains, several of which have not previously been associated witb asthma. The late expression of Chi313, Retnla and Mmp12 correlated with increased expression of IL-10 in the lungs and we hypothesise that this cytokine may be produced by alternatively activated macrophages as part of the resolution of disease. This study identifies several novel genes and mechanisms associated with the modulation of airway inflammation and pathology. The identification of factors that control allergic inflammation may provide novel therapeutic targets for disease intervention. / http://proxy.library.adelaide.edu.au/login?url= http://library.adelaide.edu.au/cgi-bin/Pwebrecon.cgi?BBID=1366239 / Thesis (Ph.D.) -- University of Adelaide, School of Molecular and Biomedical Science, 2009

Würzburg, Universiẗat, Diss., 2007.