An artificial neural network was used to optimize the release of salbutamol sulfate from hydrophilic matrix formulations. Model formulations to be used for training, testing and validating the neural network were manufactured with the aid of a central composite design with varying the levels of Methocel® K100M, xanthan gum, Carbopol® 974P and Surelease® as the input factors. In vitro dissolution time profiles at six different sampling times were used as target data in training the neural network for formulation optimization. A multi layer perceptron with one hidden layer was constructed using Matlab®, and the number of nodes in the hidden layer was optimized by trial and error to develop a model with the best predictive ability. The results revealed that a neural network with nine nodes was optimal for developing and optimizing formulations. Simulations undertaken with the training data revealed that the constructed model was useable. The optimized neural network was used for optimization of formulation with desirable release characteristics and the results indicated that there was agreement between the predicted formulation and the manufactured formulation. This work illustrates the possible utility of artificial neural networks for the optimization of pharmaceutical formulations with desirable performance characteristics.

The preparation and initiation of movement has previously been described using a neural accumulation model; this model involves an increase of neural activation in the motor cortex (M1) from baseline to a subthreshold level following a warning signal, which is maintained until presentation of an imperative stimulus (IS). Activity then increases until reaching movement initiation threshold. This model predicts that variability in activation during preparation may influence reaction time (RT) and its variability. The purpose of this thesis project was to determine whether differences in RT/variability of RT during the completion of tasks with varying levels of complexity may be attributable to differences in neural excitability in M1. To test this prediction, transcranial magnetic stimulation (TMS) delivered concurrently with an IS was used to determine neural excitability for movements with different accuracy demands. It was hypothesized that higher accuracy demands would result in lowered amplitude and/or greater variability of neural activation, and consequently slower/more variable RT. Fifteen healthy participants completed a simple RT task involving a targeted wrist extension movement under three different accuracy conditions (easy, moderate, difficult). TMS was delivered concurrently with the IS on 50% of trials during each condition. While pilot testing showed RT differences between accuracy conditions (Appendix A), the data presented here failed to detect significant differences in RT latency (F(2, 28) = .074, p = .929) or variability (F(1.432, 20.053) = .633, p = .538) between conditions . Similarly, no difference in MEP amplitude was observed between difficulty conditions (F(2, 28) = 2.439, p = .106). However, a subset of participants (n = 7) did show significant RT increases between easy and hard conditions (t(6) = 2.531, p = .045), but this subset still failed to show differences in MEP amplitude (t(6) = 1.157, p = .291) or variability (t(6) = 1.545, p = .173), suggesting that preparatory levels at the IS may be similar for movements involving both high and low accuracy demands.

Dissertação (mestrado) - Universidade Federal de Santa Catarina, Centro de Ciências Biológicas, Programa de Pós-Graduação em Biologia Celular e do Desenvolvimento, Florianópolis, 2017. / Made available in DSpace on 2017-11-07T03:25:46Z (GMT). No. of bitstreams: 1 347991.pdf: 3464947 bytes, checksum: 4b9adce4d95c51c30b0828be864fb06e (MD5) Previous issue date: 2017 / A crista neural (CN) é composta por células heterogêneas e multipotentes, e pode ser dividida em CN cefálica (CNC) e truncal (CNT). As células da CNC originam, in vivo, células neurais e mesenquimais, neste último caso, formando parte do esqueleto craniofacial (condrócitos, osteócitos e odontoblastos) e tecido conjuntivo da face (adipócitos e fibroblastos dermais). As células da CN truncal (CNT), por sua vez, originam in vivo, células gliais e neurônios do sistema nervoso periférico, além de células cromafins do sistema endócrino. Melanócitos são formados por células da CNC e CNT. In vivo, as células da CNT não apresentam a capacidade de se diferenciarem em fenótipos mesenquimais. Entretanto, in vitro, sob estímulo com fatores químicos e físicos, estas células podem se diferenciar em condrócitos, osteócitos e adipócitos. Frente a estas capacidades que as células da CNT apresentam in vitro, avaliamos a utilização de uma matriz sintética e pura chamada PuraMatrix , para cultivo e diferenciação de células da CNT. Inicialmente, testamos e padronizamos anticorpos para marcação de osteócitos (SB-1, SB-2, SB-3 e SB-5) em membros posteriores de embriões de codornas. Em seguida, realizamos culturas de células da CNT com embriões de 18-22 pares de somitos, com meio básico de cultivo e com adição de Fatores de Diferenciação Mesenquimais (FDM) para estimular as células a se diferenciarem em adipócitos e osteócitos. Foram testadas as concentrações de PuraMatrix 0,15%; 0,25%, 0,5% e 1%. Análises fenotípicas de frequência e quantitativas através da técnica de imunocitoquímica e colorações específicas foram realizadas nos 14º e 21º dias de cultivos para todos os fenótipos da CN. As quatro concentrações de PuraMatrix suportaram o cultivo de células da CNT. Entretanto, pelo baixo número de células observado, a concentração de PuraMatrix 1% foi descartada de nossas análises. As concentrações de PuraMatrix 0,15%; 0,25% e 0,5% possibilitaram a diferenciação de células da CNT para células gliais, neurônios, melanócitos, células musculares lisas e condrócitos, tanto em culturas controles quanto estimuladas com FDM. Adipócitos estiveram presentes nas três concentrações quando estimuladas com FDM, e também nas concentrações de PuraMatrix 0,15% e 0,25% na condição controle. Osteócitos foram analisados nas duas concentrações mais baixas de PuraMatrix , onde marcações para SB-3 ocorreram, mas em co-localização com cartilagem, assim como observado in vivo. SB-5 mostrou-se muito específico para osteoblastos/osteócitos e apresentou marcação tanto em cultivos controles quanto tratados. Apenas no 21º dia, houve marcação de alcalina fosfatase e de matriz mineralizada. PuraMatrix abre novas perspectivas para o desenvolvimento de estudos clonais de progenitores da CN e poderá permitir a identificação de novos progenitores com as potencialidades neurais-mesenquimais para CNT. / Abstract : The neural crest (NC) is composed of heterogenous and multipotent cells, and can be divided into cephalic (CNC) and trunkal (TNC) NC. CNC cells originate, in vivo, neural and mesenchymal cells, the last one forming part of the craniofacial skeleton (chondrocytes, osteocytes and odontoblasts) and connective tissue of the face (adipocytes and dermal fibroblasts). The TNC cells, in turn, originated in vivo, glial cells and neurons of the peripheral nervous system, in addition to chromaffin cells of the endocrine system. Melanocytes are formed by CNC and TNC cells. In vivo, TNC cells don?t have the ability to differentiate into mesenquimal phenotypes. However, in vitro, under stimulation with chemical and physical factors, these cells can differentiate into chondrocytes, osteocytes and adipocytes. Faced with these capabilities that TNC cells present in vitro, we evaluated the use of a pure synthetic matrix called PuraMatrix , for TNC cell culture and differentiation. Initially, we tested and standardized antibodies for marking osteocytes (SB-1, SB-2, SB-3 and SB-5) on hind limbs of quail embryos. Then, we performed cell cultures of TNC cells with embryos of 18-22 pairs of somites, with basic culture medium and with addition of Mesenchymal Differentiation Factors (MDF) to stimulate the cells to differentiate into adipocytes and osteocytes. The concentrations of PuraMatrix 0.15%; 0.25%, 0.5% and 1% were tested. Frequency and quantitative phenotypic analyzes using the immunocytochemistry technique and specific staining were performed on the 14th and 21st day of cultures for all NC phenotypes. The four concentrations of PuraMatrix supported the cultivation of CNT cells. However, due to the low number of cells observed, the concentration of PuraMatrix 1% was discarded from the analyzes. The concentrations of PuraMatrix 0.15%; 0.25% and 0.5% allowed the differentiation of TNC cells into glial cells, neurons, melanocytes, smooth muscle cells and chondrocytes in both control and MDF stimulated cultures. Adipocytes were present in the three concentrations when stimulated with MDF, and also in the concentrations of PuraMatrix 0.15% and 0.25% in the control condition. Osteocytes were analyzed for the two lowest concentrations of PuraMatrix , where SB-3 marker occurred with co-localization with cartilage as well as observed in vivo. SB-5 showed to be very specific for osteoblasts/osteocytes and showed labeling in both control and treated cultures. On day 21 only, there was marking of alkaline phosphatase and mineralized matrix. PuraMatrix opens new perspectives for the development of clonal studies of NC progenitors and may allow the identification of new progenitors with the neural-mesenchymal potentialities for TNC.

This thesis aims to understand how to design high performance, flexible and cost effective neural computing systems and apply them to a variety of real-time applications. Systems of this type already exist for the support of a range of ANN models. However, many of these designs have concentrated on optimising the architecture of the neural processor and have generally neglected other important aspects. If these neural systems are to be of practical benefit to researchers and allow complex neural problems to be solved efficiently, all aspects of their design must be addressed.

Modular connectionist systems comprise autonomous, communicating modules, achieving a behaviour more complex than that of a single neural network. The component modules, possibly of different topologies, may operate under various learning algorithms. Some modular connectionist systems are constrained at the representational level, in that the connectivity of the modules is hard-wired by the modeller; others are constrained at an architectural level, in that the modeller explicitly allocates each module to a specific subtask. Our approach aims to minimise these constraints, thus reducing the bias possibly introduced by the modeller. This is achieved, in the first case, through the introduction of adaptive connection weights and, in the second, by the automatic allocation of modules to subtasks as part of the learning process. The efficacy of a minimally constrained system, with respect to representation and architecture, is demonstrated by a simulation of numerical development amongst children. The modular connectionist system MASCOT (Modular Architecture for Subitising and Counting Over Time) is a dual-routed model simulating the quantification abilities of subitising and counting. A gating network learns to integrate the outputs of the two routes in determining the final output of the system. MASCOT simulates subitising through a numerosity detection system comprising modules with adaptive weights that self-organise over time. The effectiveness of MASCOT is demonstrated in that the distance effect and Fechner's law for numbers are seen to be consequences of this learning process. The automatic allocation of modules to subtasks is illustrated in a simulation of learning to count. Introducing feedback into one of two competing expert networks enables a mixture-of-experts model to perform decomposition of a task into static and temporal subtasks, and to allocate appropriate expert networks to those subtasks. MASCOT successfully performs decomposition of the counting task with a two-gated mixture-of-experts model and exhibits childlike counting errors.

M.Ing. / This dissertation discusses the results of a literature survey into the theoretical aspects and development of recurrent neural networks. In particular, the various architectures and training algorithms developed for recurrent networks are discussed. The various characteristics of importance for the efficient implementation of recurrent neural networks to model dynamical nonlinear processes have also been investigated and are discussed. Process control has been identified as a field of application where recurrent networks may play an important role. The model based adaptive control strategy is briefly introduced and the application of recurrent networks to both the direct- and the indirect adaptive control strategy highlighted. In conclusion, the important areas of future research for the successful implementation of recurrent networks in adaptive nonlinear control are identified

No description available.

This thesis is about adaptive invariance, and a new model of it: the Group Representation Network. We begin by discussing the concept of adaptive invariance. We then present standard background material, mostly from the fields of group theory and neural networks. Following this we introduce the problem of invariant pattern recognition and describe a number of methods for solving various instances of it. Next, we define the Symmetry Network, a connectionist model of permutation invariance, and we develop some new theory of this model. We also extend the applicability of the Symmetry Network to arbitrary finite group actions. We then introduce the Group Representation Network (GRN) as an abstract model, with which in principle we can construct concomitants between arbitrary group representations. We show that the GRN can be regarded as a neural network model, and that it includes the Symmetry Network as a submodel. We apply group representation theory to the analysis of GRNs. This yields general characterizations of the allowable activation functions in a GRN and of their weight matrix structure. We examine various generalizations and restricted cases of the GRN model, and in particular look at the construction of GRNs over infinite groups. We then consider the issue of a GRN's discriminability, which relates to the problem of graph isomorphism. We look next at the computational abilities of the GRN, and postulate that it is capable of approximately computing any group concomitant. We show constructively that any polynomial concomitant can be computed by a GRN. We also prove that a variety of standard models for invariant pattern recognition can be viewed as special instances of the GRN model. Finally, we propose that the GRN model may be biologically plausible and give suggestions for further research.

DNA methylation-dependent gene silencing, catalyzed by DNA methyltransferases (DNMTs) and mediated by methyl binding domain proteins (MBDs) and histone deacetylases (HDACs), is essential for mammalian development, with the nervous system demonstrating particular sensitivity to perturbations. Little is known, however, about the role of DNA methylation in the stage-specific differentiation of neurons. In the olfactory epithelium (OE), where neurogenesis is continuous and the cells demonstrate a laminar organization with a developmental hierarchy, we identified sequential, transitional stages of differentiation likely mediated by different DNMT, MBD and HDAC family members. Biochemically, HDAC1 and HDAC2 associate with repressor complexes recruited by both MBD2 and MeCP2. HDAC1 and HDAC2, however, are divergently expressed in the OE, a pattern that is recapitulated in the brain. Rather than simultaneous inclusion in a complex, therefore, the individual association of HDAC1 or HDAC2 may provide specificity to a repressor complex in different cell types. Furthermore, distinct transitional stages of differentiation are perturbed in the absence of MBD2 or MeCP2. MeCP2 is expressed in the most apical immature olfactory receptor neurons (ORNs), and is up-regulated with neuronal maturation. In the MeCP2 null OE there is a transient delay in ORN maturation and an increase in neurons of an intermediate developmental stage. Two protein variants of MBD2 are expressed in the OE, with MBD2b expressed in cycling progenitor cells and MBD2a in the maturing ORNs. MBD2 null ORNs undergo increased apoptotic cell death. There is also a significant increase in proliferating progenitors in the MBD2 null OE, likely due, at least in part, to feedback from the dying ORNs, acting to up-regulate neurogenesis. Increased cell cycling in the MBD2 null is also observed post-lesion, however, in the absence of feedback back from the ORNs, a phenotype that is recapitulated by an acute inhibition of HDACs with valproic acid. Therefore, disruptions at both transitional stages of ORN differentiation are likely in the MBD2 null mouse. Together, these results provide the first evidence for a sequential recruitment of different MBD proteins and repressor complexes at distinct transitional stages of neuronal differentiation. / Medicine, Faculty of / Graduate

Pluripotent stem cells (PSCs: ESCs and iPSCs) provide an excellent model system for studying neural development and function. These cells also serve as a reliable source of cell replacement for the treatment of various neurodegenerative diseases and disorders. In view of these applications of PSCs, multiple protocols have been developed to direct their differentiation into neural lineage. However, many of these protocols are limiting in terms of (a) low efficiency of generation of neural cells after long-term culture, (b) requirement of exogenous factors to induce and enhance neural differentiation and (c) supplementation of PSC culture medium with serum. Therefore, in the present study, attempts were made to achieve enhanced efficiency of neural differentiation of PSCs in the absence of exogenous molecules by employing a defined culture medium containing knockout serum replacement (KSR). KSR-based culture system was tested with our in-house-derived EGFP-transgenic ‘GS-2’ ES-cell and ‘N9’ iPS-cell lines and the wild-type ‘D3’ ES-cell line. In KSR medium, PSC-derived EBs predominantly generated neural cells from their post-attachment outgrowths and the complexity of neural networks increased as the culture progressed. Molecular phenotyping of PSC-derived neural cells was performed based on the expression of neural markers both at the mRNA and protein levels. qPCR analysis revealed the expression of markers corresponding to multiple neural cell types, including glutamatergic, GABAergic, cholinergic, serotonergic and dopaminergic neurons, astrocytes and oligodendrocytes, at various time points during the culture. RNA expression studies were confirmed via immunocytochemical analysis of the expression of neural markers. On day 15 of culture, FACS quantitation revealed the efficient generation of NES+ neural progenitors (~16-18%), MAP2+ mature neurons (~12-26%) and GFAP+ astrocytes (~30-63%) from the three PSC lines. Functional assessment of the generated neurons was performed by electrophysiological analysis of passive (RMP) and active (threshold, amplitude, FWHM and outward and inward currents) membrane properties. In order to investigate the role of default pathway in neural differentiation of PSCs in KSR medium, various strategies were employed. GS-2 ES-cells were cultured in the presence of different serum-free supplements; predominant differentiation into neural lineage was achieved in the B27-supplemented medium. The supplementation of KSR medium with BMP4 failed to show any effect of neural differentiation of GS-2 ES-cells. Also, EBs were switched between KSR- and FBS-supplemented culture conditions on day 2 or day 5 of culture. These experiments indicated that KSR medium promoted the generation of neural cell fates at the expense of differentiation to non-neural lineages. Interestingly, differentiation of P19 EC-cells in KSR medium also resulted in the predominant neural differentiation. These experiments collectively suggested the importance of default pathway in neural differentiation of PSCs in KSR medium. Additionally, efforts were made to enrich PSC-derived neural cells and also to enhance the efficiency of neural differentiation of PSCs. The removal of central EB-core from its peripheral neural outgrowth via scooping resulted in the enrichment of neural cells by ~1.3-2.1 folds. Significant increases were observed in the percentages of GS-2 ES-cell-derived MAP2+ mature neurons and GFAP+ astrocytes. Also, FGF2 supplementation of KSR medium was tested as a strategy to achieve enhanced efficiency of neural differentiation. Preliminary studies suggested an increase in the percentage of NES+ neural progenitors in the presence of FGF2. Taken together, KSR-based culture system offers a simple, defined and efficient method to achieve neural differentiation of PSCs in short time duration in the absence of exogenous factors. KSR-based culture system can be employed to generate specific neural cell types, study molecular regulation of neural differentiation and in disease modeling. Also, it can be used to develop a platform for high-throughput screening of potential neurogenic molecules and for dissecting their mechanisms of action.