I hypothesise that cancer cells with high aldehyde dehydrogenase (ALDHhigh) activity present a new therapeutic target and will be selectively sensitive to 5-nitrofuran pro-drugs. Cancers are heterogeneous and contain subpopulations of ALDHhigh cells with tumour initiating potential. ALDH enzymes metabolize toxic aldehydes, and are highly expressed in somatic and cancer stem cells (CSCs), although their function in CSCs is not fully understood. In a small molecule screen coupled with target ID, Zhou et al. (2012) recently discovered that clinically active 5-nitrofurans (5-NFNs) are substrates of ALDH2. 5-NFNs are a class of pro-drug widely used to treat bacterial and parasitic infections, where their relative specificity is driven by nitroreductases, but little is known about the enzymes that bio-activate 5-NFNs in humans. Recent clinical cancer research has found that the 5-NFN, nifurtimox, has anti-cancer properties and it is currently in Phase 2 clinical trials for neuroblastoma and medulloblastoma (ClinicalTrials.gov Identifier: NCT00601003), however the mechanism underlying this anti-cancer activity is unknown. In melanoma and other cancers, ALDH1A1 and ALDH1A3 are highly expressed in CSCs. I demonstrate the anti-cancer activity of 5-NFNs in cancer cell lines, where they express high sensitivity to 5-NFNs in cell viability assays (A375 melanoma cells EC50 = 867nM). To test if ALDH1 enzymes are substrates of 5-NFNs, I performed in vitro activity assays by monitoring NADH production (λ = 340nm). I found that the clinically available 5-NFNs, nifuroxazide and nifurtimox, in addition to our own newly synthesised 5-NFNs, are competitive substrates for human ALDH1A3 activity in vitro (P < 0.05). Notably, nifuroxazide is not a substrate for ALDH2, suggesting that nifuroxazide may show selectivity toward ALDH1. Enzymatic assays with purified human ALDH2, demonstrate that ALDH2 requires NAD+ for bio-activation of 5-NFNs. Consistent with these assays, I found that 5-NFNs are competitive substrates for ALDH activity in melanoma cells by Aldefluor™, with 5-NFNs displaying a prolonged competitive inhibition of ALDH activity compared with the known inhibitor, DEAB. Importantly, no-nitro control compounds show no activity toward ALDH enzymes in vitro or in culture. Kinetic living-cell imaging (IncuCyte ZOOM®) reveals that a subpopulation of ALDH1A3 siRNA transfected A375 cells are protected from 5-NFN toxicity (P > 0.05) and cell death (DRAQ7™: P < 0.0001), demonstrating a functional role for ALDH1A3 in mediating 5-NFN activity in cancer cells. In contrast, A375 cells overexpressing ALDH1A3 by cDNA transient transfection were hypersensitive to 5-NFNs (P < 0.001), determined by Muse™ cell viability. Computational docking studies reveal that 5-NFNs have the potential to fit within the interior of the ALDH enzymatic cavity and interact with the catalytic cysteine, thereby offering a potential mechanism for 5-NFN bio-activation. Finally, in collaboration, we show a unique interaction between 5-NFNs and ALDH using mass spectrometry and have initiated protein crystallography trials. My work demonstrates a novel and biologically relevant 5-NFN-ALDH interaction in cancer cells. I propose 5-NFNs have the potential to target ALDHhigh CSCs within a tumour and advance the repurposing of clinical 5-NFN pro-drug antibiotics as anti-cancer therapeutics.

nÃo hà / Sistemas climatolÃgicos sÃo caracterizados por apresentarem modelagem complexa e de baixa previsibilidade. Em regiÃes de clima semiÃrido, como o Nordeste Brasileiro, informaÃÃes de previsÃo climatolÃgicas sÃo de interesse para um melhor aproveitamento dos recursos hÃdricos. O Estado do CearÃ, localizado no norte do Nordeste Brasileiro, sofre periodicamente com os problemas de estiagem. Atualmente a FundaÃÃo Cearense de Meteorologia e Recursos HÃdricos (FUNCEME), ÃrgÃo pertencente ao governo do Estado do CearÃ, à responsÃvel por gerar pesquisas voltadas a trazer um melhor entendimento fenomenolÃgico do clima do Estado e com isso efetuar uma melhor previsÃo de como serà o perÃodo de chuvas. Hoje a FundaÃÃo utiliza-se de modelagem numÃrica composta por dois modelos regionais, Modelo Regional Espectral 97 (MRE) e o Regional Modeling Atmospheric System (RAMS), aninhados por uma tÃcnica de downscaling ao modelo dinÃmico de grande escala ECHAM4.5, para efetuar suas previsÃes. Os modelos dinÃmicos sÃo caracterizados por apresentarem elevado custo computacional, grande quantidade de dados para sua entrada e alta complexidade na utilizaÃÃo. O desenvolvimento de modelos de previsÃo baseados em Redes Neurais Artificias (RNA) abrange diversas Ãreas do conhecimento e tem apresentado resultados promissores. Modelos baseados em redes neurais sÃo capazes de reproduzir deferentes tipos de sistemas atravÃs da sua capacidade de aprendizado. Nesta dissertaÃÃo foi desenvolvido um modelo de previsÃo de chuvas para as oito regiÃes homogÃneas do Estado do CearÃ, que apresenta um baixo custo computacional e de fÃcil utilizaÃÃo. Para atingir este desenvolvimento foi utilizada uma RNA baseada na tÃcnica Neo-Fuzzy Neuron (NFN). Apesar de ser proposto um novo modelo de previsÃo, nÃo se deseja a substituiÃÃo dos atuais modelos, o novo modelo proposto nesta dissertaÃÃo tem por finalidade enriquecer as informaÃÃes geradas atravÃs de modelos de previsÃo para que assim possa ser gerada uma melhor prediÃÃo de como serà o perÃodo de chuvas no Estado do CearÃ. O modelo proposto foi comparado ao modelo MRE que à atualmente utilizado pela FUNCEME para suas previsÃes. Nesta comparaÃÃo utilizou-se como indicadores de desempenho: tempo de execuÃÃo, valor da raiz quadrada do erro mÃdio quadrÃtico (REMQ) e a correlaÃÃo com os valores observados. Ao final pode-se concluir que o modelo desenvolvido apresentou um melhor desempenho com menor tempo de processamento em relaÃÃo ao modelo dinÃmico MRE para efetuar a previsÃo de chuvas. / Climatological systems are characterized by complex modeling and having low predictability. In semi-arid regions, as the Brazilian Northeast, weather forecast information are necessary for the maintenance of life and a better use of water resources. The State of CearÃ, located on the north of Brazilian Northeast, is a region that suffers with drought for a long time. The FundaÃÃo Cearense de Meteorologia e Recursos HÃdricos (FUNCEME), which belongs to the state government, is responsible for generating research to bring a better phenomenological understanding on the weather of the State of Cearà and thus make a better prediction on how the rainy season will be. Today the foundation makes use of numerical modeling consisting of two regional models, the Regional Spectral Model (RSM) and the Regional Modeling Atmospheric System (RAMS), nested by a downscaling technique to the large scale dynamic model ECHAM4.5, in order to do its predictions. Dynamic models are characterized by their high computational costs, large amounts of information on its input and high complexity usage. The development of forecasting models based on Artificial Neural Networks (ANN) covers various areas of knowledge showing promising results. Neural network based models are capable of reproducing different types of systems through its learning capability. In this thesis it was developed a model for predicting rain for the eight homogeneous regions of the state of Cearà that presents low computational cost and easy use. In order to achieve this development it was used an ANN base on a Neo-Fuzzy Neuron (NFN) technique. Despite being offered a new prediction model, this thesis aims to enrich the information generated by forecast models and do a better prediction on the rainy season of the State of CearÃ. The proposed model was compared to the RSM model that is currently in use by FUNCEME in its predictions. In this comparison, as performance indicators, it was used: the execution time, value of the root mean square error (RMSE) and the correlation with the observed values. At the end, it is concluded that the proposed model had a better performance and was faster than the RSM dynamic model in its predictions.

This study presents the implementation of Middle East Technical University Neural Network (METU-NN) models for the ionospheric forecasting together with worldwide usage capability of the Internet. Furthermore, an attempt is made to include expert information in the Neural Network (NN) model in the form of neurofuzzy network (NFN). Middle East Technical University Neurofuzzy Network (METU-NFN) modeling approach is developed which is the first attempt of using a neurofuzzy model in the ionospheric forecasting studies. The Web based applications developed in this study have the ability to be customized such that other NN and NFN models including METU-NFN can also be adapted. The NFN models developed in this study are compared with the previously developed and matured METU-NN models. At this very early stage of employing neurofuzzy models in this field, ambitious objectives are not aimed. Applicability of the neurofuzzy systems on the ionospheric forecasting studies is only demonstrated. Training and operating METU-NN and METU-NFN models under equal conditions and with the same data sets, the cross correlation of obtained and measured values are 0.9870 and 0.9086 and the root mean square error (RMSE) values of 1.7425 TECU and 4.7987 TECU are found by operating METU-NN and METU-NFN models respectively. The results obtained by METU-NFN model is close to those found by METU-NN model. These results are reasonable enough to encourage further studies on neurofuzzy models to benefit from expert information. Availability of these models which already attracted intense international attention will greatly help the related scientific circles to use the models. The models can be architecturally constructed, trained and operated on-line. To the best of our knowledge this is the first application that gives the ability of on-line model usage with these features. Applicability of NFN models to the ionospheric forecasting is demonstrated. Having ample flexibility the constructed model enables further developments and improvements. Other neurofuzzy systems in the literature might also lead to better achievements.