A Hybrid Forward Algorithm for RBF Neural Network Construction

This paper proposes a novel hybrid forward algorithm (HFA) for the construction of radial basis function (RBF) neural networks with tunable nodes. The main objective is to efficiently and effectively produce a parsimonious RBF neural network that generalizes well. In this study, it is achieved through simultaneous network structure determination and parameter optimization on the continuous parameter space. This is a mixed integer hard problem and the proposed HFA tackles this problem using an integrated analytic framework, leading to significantly improved network performance and reduced memory usage for the network construction. The computational complexity analysis confirms the efficiency of the proposed algorithm, and the simulation results demonstrate its effectiveness

Ruthenium complexes of the 1,4-bis(diphenylphosphino)-1,3-butadiene-bridged diphosphine 1,2,3,4-Me-4-NUPHOS: Solvent-dependent interconversion of four- and six-electron donor coordination and transfer hydrogenation activity

In chloroform, [RuCl2(nbd)(py)(2)] (1) (nbd = norbornadiene; py = pyridine) reacts with 1,4-bis(diphenylphosphino)-1,2,3,4-tetramethyl-1,3-butadiene (1,2,3,4-Me-4-NUPHOS) to give the dimer [Ru2Cl3(eta(4)-1,2,3,4-Me-4-NUPHOS)(2)]Cl (2a), whereas, in THF [RuCl2(1,2,3,4-Me-4-NUPHOS)(PY)(2)] (3) is isolated as the sole product of reaction. Compound 2 exists as a 4:1 mixture of two noninterconverting isomers, the major with C, symmetry and the minor with either C, or C-2 symmetry. A single-crystal X-ray analysis of [Ru2Cl3 (eta(4)-1,2,3,4-Me-4-NUPHOS)(2)] [SbF6] (2b), the hexafluoroantimonate salt of 2a, revealed that the diphosphine coordinates in an unusual manner, as a eta(4)-six-electron donor, bonded through both P atoms and one of the double bonds of the butadiene tether. Compounds 2a and 3 react with 1,2-ethylenediamine (en) in THF to afford [RuCl2(1,2,3,4-Me-4-NUPHOS)(en)] (4), which rapidly dissociates a chloride ligand in chloroform to give [RuCl(eta(4)-1,2,3,4-Me-4-NUPHOS)(en)] [Cl] (5a). Complexes 4 and 5a cleanly and quantitatively interconvert in a solvent-dependent equilibrium, and in THF 5a readily adds chloride to displace the eta(2)-interaction and re-form 4. A single-crystal X-ray structure determination of [RuCl(eta(4)-1,2,3,4-Me-4-NUPHOS)(en)][ClO4] (5b) confirmed that the diphosphine coordinates in an eta(4)-manner as a facial six-electron donor with the eta(2)-coordinated double bond occupying the site trans to chloride. The eta(4)-bonding mode can be readily identified by the unusually high-field chemical shift associated with the phosphorus atom adjacent to the eta(2)-coordinated double bond. Complexes 2a, 2b, 4, and 5a form catalysts that are active for transfer hydrogenation of a range of ketones. In all cases, catalysts formed from precursors 2a and 2b are markedly more active than those formed from 4 and 5a.

A ring-closing metathesis approach to the bicyclo[4.3.1]decane core of caryolanes

A synthesis of highly substituted and sterically congested bicyclo[4.3.1]decenes, a structure embedded in the core 4,7,6-tricyclic system of natural caryolanes, was successfully achieved via a ring-closing metathesis (RCM) reaction of syn-1,3-diene substituted cyclohexanols. The construction of the diene substrates, starting from 4-acetoxy-3-methyl-2-cyclohexen-1-one, employed diastereoselective copper-mediated conjugate addition and Grignard reactions. An X-ray crystal structure determination of a key synthetic intermediate confirmed the relative stereochemistry of the RCM bicyclic product.

Two-stage mixed discrete-continuous identification of radial basis function (RBF) neural models for nonlinear systems

The identification of nonlinear dynamic systems using radial basis function (RBF) neural models is studied in this paper. Given a model selection criterion, the main objective is to effectively and efficiently build a parsimonious compact neural model that generalizes well over unseen data. This is achieved by simultaneous model structure selection and optimization of the parameters over the continuous parameter space. It is a mixed-integer hard problem, and a unified analytic framework is proposed to enable an effective and efficient two-stage mixed discrete-continuous; identification procedure. This novel framework combines the advantages of an iterative discrete two-stage subset selection technique for model structure determination and the calculus-based continuous optimization of the model parameters. Computational complexity analysis and simulation studies confirm the efficacy of the proposed algorithm.

Synthesis and characterisation of new platinum-acetylide complexes containing diimine ligands

A new class of platinum-bipyridyl compounds has been synthesized by the dehydrohalogenative reaction of [4,4'-bis(tert-butyl)-2,2'-bipyridyl]platinum dichloride [PtCl2((t)Bu(2)bipy)] 1 with terminal alkynes HC=CR, in the presence of copper(I) iodide and diisopropylamine. The products [Pt(C=CR)(2)((t)Bu(2)bipy)] (R=C6H4NO2-p 2, C6H5 3, C6H4CH3-p 4 or SiMe3 5), have been characterised by spectroscopic and analytical methods, and a single crystal molecular structure determination has been carried out on 4. Extended Huckel molecular orbital calculations have also been carried out, and the results are used to help rationalise the voltammetric, EPR and spectroelectrochemical properties of the new compounds. These show that compounds 3, 4 and 5 undergo a one-electron bipyridyl based redox process, but that 2 has an unresolved two-electron process located on the nitro groups.

Polymorphism in Sulfadimidine/4-Aminosalicylic Acid Cocrystals: Solid-State Characterization and Physicochemical Properties

Polymorphism of crystalline drugs is a common phenomenon. However, the number of reported polymorphic cocrystals is very limited. In this work, the synthesis and solid-state characterization of a polymorphic cocrystal composed of sulfadimidine (SD) and 4-aminosalicylic acid (4-ASA) is reported for the first time. By liquid-assisted milling, the SD:4-ASA 1:1 form I cocrystal, the structure of which has been previously reported, was formed. By spray drying, a new polymorphic form (form II) of the SD:4-ASA 1:1 cocrystal was discovered which could also be obtained by solvent evaporation from ethanol and acetone. Structure determination of the form II cocrystal was calculated using high-resolution X-ray powder diffraction. The solubility of the SD:4-ASA 1:1 cocrystal was dependent on the pH and predicted by a model established for a two amphoteric component cocrystal. The form I cocrystal was found to be thermodynamically more stable in aqueous solution than form II, which showed transformation to form I. Dissolution studies revealed that the dissolution rate of SD from both cocrystals was enhanced when compared with a physical equimolar mixture and pure SD.

Design of neuro-fuzzy models by evolutionary and gradient-based algorithms

All systems found in nature exhibit, with different degrees, a nonlinear behavior. To emulate this behavior, classical systems identification techniques use, typically, linear models, for mathematical simplicity. Models inspired by biological principles (artificial neural networks) and linguistically motivated (fuzzy systems), due to their universal approximation property, are becoming alternatives to classical mathematical models. In systems identification, the design of this type of models is an iterative process, requiring, among other steps, the need to identify the model structure, as well as the estimation of the model parameters. This thesis addresses the applicability of gradient-basis algorithms for the parameter estimation phase, and the use of evolutionary algorithms for model structure selection, for the design of neuro-fuzzy systems, i.e., models that offer the transparency property found in fuzzy systems, but use, for their design, algorithms introduced in the context of neural networks. A new methodology, based on the minimization of the integral of the error, and exploiting the parameter separability property typically found in neuro-fuzzy systems, is proposed for parameter estimation. A recent evolutionary technique (bacterial algorithms), based on the natural phenomenon of microbial evolution, is combined with genetic programming, and the resulting algorithm, bacterial programming, advocated for structure determination. Different versions of this evolutionary technique are combined with gradient-based algorithms, solving problems found in fuzzy and neuro-fuzzy design, namely incorporation of a-priori knowledge, gradient algorithms initialization and model complexity reduction.

Études structurales d’interactions protéine/protéine impliquées dans l’érythropoïèse

Le développement hématopoïétique est régulé par l’action combinée de facteurs de transcription lignée spécifiques et de la machinerie transcriptionnelle de base, permettant ainsi l’expression de gènes en temps et lieu appropriés. Les travaux présentés dans cette thèse portent sur l’étude structurale et fonctionnelle d’interactions décisives pour la régulation de l’expression de gènes et impliquant des domaines de transactivation (TAD). En effet, les interactions faisant intervenir les TAD d’activateurs permettent de réguler l’activation de la transcription de façon spécifique. La première étude présentée dans cette thèse relate l'identification et la caractérisation d'une nouvelle interaction entre deux facteurs de transcription : le facteur hématopoïétique GATA-1 et la protéine suppresseur de tumeur p53. En combinant des études in vitro par titrage calorimétrique en condition isotherme (ITC) et par spectroscopie RMN et des études in vivo, nous avons identifié et caractérisé cette nouvelle interaction. Il s'avère que le TAD de p53 et le domaine de liaison à l’ADN de GATA-1 sont les domaines minimaux requis pour la formation de ce complexe. L'inhibition de la voie p53 par GATA-1 s’est avérée être la conséquence majeure de cette interaction, permettant ainsi le maintien en vie des précurseurs érythrocytaires via l’inhibition de l’apoptose. Un deuxième type d’interaction a fait l’objet d’études : l’interaction entre divers TAD et la machinerie transcriptionnelle de base, plus spécifiquement avec le Facteur général de Transcription IIH (TFIIH). La structure des complexes constitués par la sous-unité Tfb1/p62 du facteur TFIIH en interaction avec le TAD viral de VP16 d’une part, et avec le TAD humain du facteur érythrocytaire « Erythroid Krüppel-like factor» (EKLF) d’autre part, ont été résolues par spectroscopie RMN. La structure du complexe Tfb1/VP16 a révélée que le mode de liaison de VP16 à Tfb1 est similaire au mode de liaison du TAD de p53 avec le même partenaire. En effet, les TAD de VP16 et de p53 forment tous deux une hélice α de 9 résidus en interaction avec Tfb1. En dépit de partager avec p53 et VP16 le même site de liaison sur Tfb1/p62, la structure RMN du complexe EKLF/Tfb1 démontre que le mode d’interaction de ce TAD se distingue du mode de liaison canonique des activeurs transcriptionnels. Etonnamment, EKLF adopte un mécanisme de liaison semblable au mécanisme de liaison du facteur général de transcription TFIIEα avec p62, leurs conformations demeurent étendues en interaction avec Tfb1/p62. En se basant sur nos données structurales, nous avons identifié un résidu dans le TAD d'EKLF décisif pour la formation du complexe EKLF/p62 : le Trp73. La mutation de cet acide aminé perturbe son interaction avec Tfb1PH/p62PH et réduit significativement l'activité transcriptionnelle d'EKLF dans les érythrocytes.

Caractérisation structurale et thermodynamique de la reconnaissance du substrat par le ribozyme VS de Neurospora

Les interactions ARN/ARN de type kissing-loop sont des éléments de structure tertiaire qui jouent souvent des rôles clés chez les ARN, tant au niveau fonctionnel que structural. En effet, ce type d’interaction est crucial pour plusieurs processus dépendant des ARN, notamment pour l’initiation de la traduction, la reconnaissance des ARN antisens et la dimérisation de génome rétroviral. Les interactions kissing-loop sont également importantes pour le repliement des ARN, puisqu’elles permettent d’établir des contacts à longue distance entre différents ARN ou encore entre les domaines éloignés d’un même ARN. Ce type d’interaction stabilise aussi les structures complexes des ARN fonctionnels tels que les ARNt, les riborégulateurs et les ribozymes. Comme d’autres ARN fonctionnels, le ribozyme VS de Neurospora contient une interaction kissing-loop importante. Celle-ci est impliquée dans la reconnaissance du substrat et se forme entre la tige-boucle I (stem-loop I, SLI) du substrat et la tige-boucle V (stem-loop V, SLV) du domaine catalytique. Des études biochimiques ont démontré que l’interaction kissing-loop I/V, dépendante du magnésium, implique trois paires de bases Watson-Crick (W-C). De plus, cette interaction est associée à un réarrangement de la structure du substrat, le faisant passer d’une conformation inactive dite unshifted à une conformation active dite shifted. Les travaux présentés dans cette thèse consistent en une caractérisation structurale et thermodynamique de l’interaction kissing-loop I/V du ribozyme VS, laquelle est formée de fragments d’ARN représentant les tige-boucles I et V dérivées du ribozyme VS (SLI et SLV). Cette caractérisation a été réalisée principalement par spectroscopie de résonance magnétique nucléaire (RMN) et par titrage calorimétrique isotherme (isothermal titration calorimetry, ITC) en utilisant différents complexes SLI/SLV dans lesquels l’ARN SLV est commun à tous les complexes, alors que différentes variations de l’ARN SLI ont été utilisées, soit en conformation shiftable ou preshifted. Les données d’ITC ont permis de démontrer qu’en présence d’une concentration saturante de magnésium, l’affinité d’un substrat SLI preshifted pour SLV est extrêmement élevée, rendant cette interaction plus stable que ce qui est prédit pour un duplexe d’ARN équivalent. De plus, l’étude effectuée par ITC montre que des ARN SLI preshifted présentent une meilleure affinité pour SLV que des ARN SLI shiftable, ce qui a permis de calculer le coût énergétique associé au réarrangement de structure du substrat. En plus de confirmer la formation des trois paires de bases W-C prédites à la jonction I/V, les études de RMN ont permis d’obtenir une preuve structurale directe du réarrangement structural des substrats SLI shiftable en présence de magnésium et de l’ARN SLV. La structure RMN d’un complexe SLI/SLV de grande affinité démontre que les boucles terminales de SLI et SLV forment chacune un motif U-turn, ce qui facilite l’appariement W-C intermoléculaire. Plusieurs autres interactions ont été définies à l’interface I/V, notamment des triplets de bases, ainsi que des empilements de bases. Ces interactions contribuent d’ailleurs à la création d’une structure présentant un empilement continu, c’est-à-dire qui se propage du centre de l’interaction jusqu’aux bouts des tiges de SLI et SLV. Ces études de RMN permettent donc de mieux comprendre la stabilité exceptionnelle de l’interaction kissing-loop I/V au niveau structural et mènent à l’élaboration d’un modèle cinétique de l’activation du substrat par le ribozyme VS. En considérant l’ensemble des données d’ITC et de RMN, l’étonnante stabilité de l’interaction I/V s’explique probablement par une combinaison de facteurs, dont les motifs U-turn, la présence d’un nucléotide exclu de la boucle de SLV (U700), la liaison de cations magnésium et l’empilement de bases continu à la jonction I/V.

Microbial relatives of seed storage proteins: conservation of motifs in a functionally diverse superfamily of enzymes

Plant storage proteins comprise a major part of the human diet. Sequence analysis has revealed that these proteins probably share a common ancestor with a fungal oxalate decarboxylase and/or related bacterial genes. Additionally, all these proteins share a central core sequence with several other functionally diverse enzymes and binding proteins, many of which are associated with synthesis of the extracellular matrix during sporulation/encystment. A possible prokaryotic relative of this sequence is a bacterial protein (SASP) known to bind to DNA and thereby protect spores from extreme environmental conditions. This ability to maintain cell viability during periods of dehydration in spores and seeds may relate to absolute conservation of residues involved in structure determination.

Overproduction, purification and preliminary X-ray diffraction analysis of YncE, an iron-regulated Sec-dependent periplasmic protein from Escherichia coli

The yncE gene of Escherichia coli encodes a predicted periplasmic protein of unknown function. The gene is de-repressed under iron restriction through the action of the global iron regulator Fur. This suggests a role in iron acquisition, which is supported by the presence of the adjacent yncD gene encoding a potential TonB-dependent outer-membrane transporter. Here, the preliminary crystallographic structure of YncE is reported, revealing that it consists of a seven-bladed beta-propeller which resembles the corresponding domain of the `surface-layer protein' of Methanosarcina mazei. A full structure determination is under way in order to provide insight into the function of this protein.

Preliminary X-ray diffraction analysis of YcdB from Escherichia coli: a novel haem-containing and Tat-secreted periplasmic protein with a potential role in iron transport

YcdB is a periplasmic haem-containing protein from Escherichia coli that has a potential role in iron transport. It is currently the only reported haem-containing Tat-secreted substrate. Here, the overexpression, purification, crystallization and structure determination at 2.0 angstrom resolution are reported for the apo form of the protein. The apo-YcdB structure resembles those of members of the haem-dependent peroxidase family and thus confirms that YcdB is also a member of this family. Haem-soaking experiments with preformed apo-YcdB crystals have been optimized to successfully generate haem-containing YcdB crystals that diffract to 2.9 angstrom. Completion of model building and structure refinement are under way.

Eukaryotic expression: developments for structural proteomics

The production of sufficient quantities of protein is an essential prelude to a structure determination, but for many viral and human proteins this cannot be achieved using prokaryotic expression systems. Groups in the Structural Proteomics In Europe ( SPINE) consortium have developed and implemented high- throughput ( HTP) methodologies for cloning, expression screening and protein production in eukaryotic systems. Studies focused on three systems: yeast ( Pichia pastoris and Saccharomyces cerevisiae), baculovirusinfected insect cells and transient expression in mammalian cells. Suitable vectors for HTP cloning are described and results from their use in expression screening and protein-production pipelines are reported. Strategies for coexpression, selenomethionine labelling ( in all three eukaryotic systems) and control of glycosylation ( for secreted proteins in mammalian cells) are assessed.

Bending, twisting, and breaking CuCN chains to produce framework materials: the reactions of CuCN with alkali-metal halides

The reactions of the low-temperature polymorph of copper(I) cyanide (LT-CuCN) with concentrated aqueous alkali-metal halide solutions have been investigated. At room temperature, KX (X = Br and I) and CsX (X = Cl, Br, and I) produce the addition products K[Cu-2(CN)(2)Br](H2O)-H-. (I), K-3[Cu-6(CN)(6)I-3](.)2H(2)O (II), Cs[Cu-3(CN)(3)Cl] (III), Cs[Cu-3(CN)(3)Br] (IV), and Cs-2[Cu-4(CN)(4)I-2](H2O)-H-. (V), with 3-D frameworks in which the -(CuCN)- chains present in CuCN persist. No reaction occurs, however, with NaX (X = Cl, Br, I) or KCl. The addition compounds, I-V, reconvert to CuCN when washed. Both low- and high-temperature polymorphs of CuCN (LT- and HT-CuCN) are produced, except in the case of Cs[Cu-3(CN)(3)Cl] (III), which converts only to LT-CuCN. Heating similar AX-CuCN reaction mixtures under hydrothermal conditions at 453 K for 1 day produces single crystals of I-V suitable for structure determination. Under these more forcing conditions, reactions also occur with NaX (X = Cl, Br, I) and KCl. NaBr and KCl cause some conversion of LT-CuCN into HT-CuCN, while NaCl and NaI, respectively, react to form the mixed-valence Cu(I)/Cu(II) compounds [Cu-II(OH2)(4)][Cu-4(I)(CN)(6)], a known phase, and [Cu-II(OH2)(4)][Cu-4(I)(CN)(4)I-2] (VI), a 3-D framework, which contains infinite -(CuCN)- chains. After 3 days of heating under hydrothermal conditions, the reaction between KI and CuCN produces [Cu-II(OH2)(4)][Cu-2(I)(CN)I-2](2) (VII), in which the CuCN chains are broken into single Cu-CN-Cu units, which in turn are linked into chains via iodine atoms and then into layers via long Cu-C and Cu-Cu interactions.

Separation of Ba-137m from Cs-137 using new ion exchanger Na-2(H2O)(4)(H3O)[Al(OH)(6)Mo6O18]

A new polyoxometalate of chemical formula, Na-2(H2O)(4)(H3O)[Al(OH)(6)Mo6O18] (1) containing Anderson type large anion has been synthesized and characterized by single-crystal X-ray structure determination and IR spectroscopic studies. The crystal of 1 is triclinic, spacegroup P-1 with cell dimensions, a = 6.365(9) angstrom, b = 10.37(1) angstrom, c = 10.44(1) angstrom and alpha = 65.41(1), beta = 77.18(1), gamma = 86.58(1) and Z = 1. The compound 1 behaves as an ion exchanger and is stable in thermal, radiation and chemical environments. Radiochemical separation of the short-lived daughter Ba-137m (t(1/2) = 2.50 min) from its long-lived parent Cs-137 using this newly designed and synthesized ion exchanger has been developed.