Short-term prediction of chaotic time series by using RBF network with regression weights

We propose a framework for constructing and training a radial basis function (RBF) neural network. The structure of the gaussian functions is modified using a pseudo-gaussian function (PG) in which two scaling parameters sigma are introduced, which eliminates the symmetry restriction and provides the neurons in the hidden layer with greater flexibility with respect to function approximation. We propose a modified PG-BF (pseudo-gaussian basis function) network in which the regression weights are used to replace the constant weights in the output layer. For this purpose, a sequential learning algorithm is presented to adapt the structure of the network, in which it is possible to create a new hidden unit and also to detect and remove inactive units. A salient feature of the network systems is that the method used for calculating the overall output is the weighted average of the output associated with each receptive field. The superior performance of the proposed PG-BF system over the standard RBF are illustrated using the problem of short-term prediction of chaotic time series.     

Development of simple fitness landscapes for peptides by artificial neural filter systems

The applicability of artificial neural filter systems as fitness functions for sequence-oriented peptide design was evaluated. Two example applications were selected: classification of dipeptides according to their hydrophobicity and classification of proteolytic cleavage-sites of protein precursor sequences according to their mean hydrophobicities and mean side-chain volumes. The cleavage-sites covered 12 residues. In the dipeptide experiments the objective was to separate a selected set of molecules from all other possible dipeptide sequences. Perceptrons, feedforward networks with one hidden layer, and a hybrid network were applied. The filters were trained by a (1,) evolution strategy. Two types of network units employing either a sigmoidal or a unimodal transfer function were used in the feedforward filters, and their influence on classification was investigated. The two-layer hybrid network employed gaussian activation functions. To analyze classification of the different filter systems, their output was plotted in the two-dimensional sequence space. The diagrams were interpreted as fitness landscapes qualifying the markedness of a characteristic peptide feature which can be used as a guide through sequence space for rational peptide design. It is demonstrated that the applicability of neural filter systems as a heuristic method for sequence optimization depends on both the appropriate network architecture and selection of representative sequence data. The networks with unimodal activation functions and the hybrid networks both led to a number of local optima. However, the hybrid networks produced the best prediction results. In contrast, the filters with sigmoidal activation produced good reclassification results leading to fitness landscapes lacking unreasonable local optima. Similar results were obtained for classification of both dipeptides and cleavage-site sequences.     

The perception of a dotted line in noise: a model of good continuation and some experimental results

A formal model is proposed, describing how the perceptual interpretation of dot figures is guided by the Gestalt rule of good continuation. The algorithm will be restricted to figures with a collinear dot array (line) embedded in a background of randomly placed dots. The model, CODE-2, is an elaboration of the model, CODE-1, of grouping dots on the basis of the Gestalt rule of (relative) proximity, and consists of the introduction of non-circular symmetric gaussian distribution functions for the representation of the orientation dependent strength of interaction between collinear dots. Supra-threshold contours of the function, resulting from a superposition on each dot of the gaussian functions, are assumed to predict the perceptual grouping of the dots. A quantitative measure for the perceptual salience of dotted lines was defined as the contrast between the internal coherence of the line dots, and their interference with the noise dots. For 20 stimuli the CODE-2 grouping of the dots is reported, together with the results of a line-in-noise latency experiment. There was a significant correlation between the predicted saliences and the experimental results. The results support the usefulness of representing good continuation between collinear dots by non-circular symmetric gaussian distribution functions.     

The Frequency Response, Coherence, and Information Capacity of Two Neuronal Models

Two neuronal models are analyzed in which subthreshold inputs are integrated either without loss (perfect integrator) or with a decay which follows an exponential time course (leaky integrator). Linear frequency response functions for these models are compared using sinusoids, Poisson-distributed impulses, or gaussian white noise as inputs. The responses of both models show the nonlinear behavior characteristic of a rectifier for sinusoidal inputs of sufficient amplitude. The leaky integrator shows another nonlinearity in which responses become phase locked to cyclic stimuli. Addition of white noise reduces the distortions due to phase locking. Both models also show selective attenuation of high-frequency components with white noise inputs. Input, output, and cross-spectra are computed using inputs having a broad frequency spectrum. Measures of the coherence and information transmission between the input and output of the models are also derived. Steady inputs, which produce a constant “carrier” rate, and intrinsic sources, which produce variability in the discharge of neurons, may either increase or decrease coherence; however, information transmission using inputs with a broad spectrum is generally increased by steady inputs and reduced by intrinsic variability.     

Successful control of viruses by NK cells--a balance of opposing forces?

Natural killer (NK) cells play a crucial role in limiting the severity of diseases caused by a range of viruses. Recent data have shown that the effector functions of NK cells can be specifically stimulated when NK cell activation receptors engage cellular major histocompatibility complex (MHC) class I-like ligands induced after infection or by specific viral gene products. However, to counter this NK cell response viruses have evolved an array of strategies to subvert efficient NK cell activation. These data indicate that the balance of host NK cell responses and viral NK cell escape mechanisms can be strategically poised as each strives for survival.     

An on-line realtime cardiotachograph]

A realtime cardiotachogram was devised to detect fluctuation of cardiac rhythm. The apparatus is composed of five parts; (1) a "preamplifier" for recording electrical and/or mechanical cardiac activities, (2) a "slicer" to obtain sampling pulses from cardiac activity at a certain trigger level, (3) a "trigger pulse generator" to monitor sampling pulse, (4) an "oscillator" whose output is fed to a pulse counter, and (5) a "pulse counter" which counts the outputs of the oscillator during the gating period determined by the sampling pulses. The count numbers are converted to analog output, thus we can get sawtooth wave whose amplitude is directly proportional to the cardiac interval. The cardiac intervals between 100 ms and 5 seconds can be recorded by this tachograph.     

A new method for testing gaussian processes

A method of testing gaussian processes, based on the characteristic function, is presented. An evaluation of possible gaussian inclusions in some stochastic processes is also proposed.     

Digital Counter Tallys Number of Tissue Sections Cut and Recycles at Selectable Intervals

A digital counter is described for use on any microtome. The counter performs three independent functions: 1) a cumulative count is maintained, 2) a second interval counter recycles each time a preselected number of sections are cut, and 3) a third rounds counter records the number of times the interval counter recycles and also recycles at selected intervals. The digital counter activates relays when recycling and can control motorized microtomes during serial sectioning.     

Seven sirtuins for seven deadly diseases ofaging

Sirtuins are a class of NAD⁺-dependent deacetylases having beneficial health effects. This extensive review describes the numerous intracellular actions of the seven mammalian sirtuins, their protein targets, intracellular localization, the pathways they modulate, and their role in common diseases of aging. Selective pharmacological targeting of sirtuins is of current interest in helping to alleviate global disease burden. Since all sirtuins are activated by NAD⁺, strategies that boost NAD⁺ in cells are of interest. While most is known about SIRT1, the functions of the six other sirtuins are now emerging. Best known is the involvement of sirtuins in helping cells adapt energy output to match energy requirements. SIRT1 and some of the other sirtuins enhance fat metabolism and modulate mitochondrial respiration to optimize energy harvesting. The AMP kinase/SIRT1–PGC-1α–PPAR axis and mitochondrial sirtuins appear pivotal to maintaining mitochondrial function. Downregulation with aging explains much of the pathophysiology that accumulates with aging. Posttranslational modifications of sirtuins and their substrates affect specificity. Although SIRT1 activation seems not to affect life span, activation of some of the other sirtuins might. Since sirtuins are crucial to pathways that counter the decline in health that accompanies aging, pharmacological agents that boost sirtuin activity have clinical potential in treatment of diabetes, cardiovascular disease, dementia, osteoporosis, arthritis, and other conditions. In cancer, however, SIRT1 inhibitors could have therapeutic value. Nutraceuticals such as resveratrol have a multiplicity of actions besides sirtuin activation. Their net health benefit and relative safety may have originated from the ability of animals to survive environmental changes by utilizing these stress resistance chemicals in the diet during evolution. Each sirtuin forms a key hub to the intracellular pathways affected.     

HAMP domain structural determinants for signalling and sensory adaptation in Tsr,the Escherichia coli serine chemoreceptor

HAMP domains mediate input–output transactions in many bacterial signalling proteins. To clarify the mechanistic logic of HAMP signalling, we constructed Tsr‐HAMP deletion derivatives and characterized their steady‐state signal outputs and sensory adaptation properties with flagellar rotation and receptor methylation assays. Tsr molecules lacking the entire HAMP domain or just the HAMP‐AS2 helix generated clockwise output signals, confirming that kinase activation is the default output state of the chemoreceptor signalling domain and that attractant stimuli shift HAMP to an overriding kinase‐off signalling state to elicit counter‐clockwise flagellar responses. Receptors with deletions of the AS1 helices, which free the AS2 helices from bundle‐packing constraints, exhibited kinase‐off signalling behaviour that depended on three C‐terminal hydrophobic residues of AS2. We conclude that AS2/AS2′ packing interactions alone can play an important role in controlling output kinase activity. Neither kinase‐on nor kinase‐off HAMP deletion outputs responded to sensory adaptation control, implying that out‐of‐range conformations or bundle‐packing stabilities of their methylation helices prevent substrate recognition by the adaptation enzymes. These observations support the previously proposed biphasic, dynamic‐bundle mechanism of HAMP signalling and additionally show that the structural interplay of helix‐packing interactions between HAMP and the adjoining methylation helices is critical for sensory adaptation control of receptor output.     

Automatic determination of radial basis functions: an immunity-based approach

The appropriate operation of a radial basis function (RBF) neural network depends mainly upon an adequate choice of the parameters of its basis functions. The simplest approach to train an RBF network is to assume fixed radial basis functions defining the activation of the hidden units. Once the RBF parameters are fixed, the optimal set of output weights can be determined straightforwardly by using a linear least squares algorithm, which generally means reduction in the learning time as compared to the determination of all RBF network parameters using supervised learning. The main drawback of this strategy is the requirement of an efficient algorithm to determine the number, position, and dispersion of the RBFs. The approach proposed here is inspired by models derived from the vertebrate immune system, that will be shown to perform unsupervised cluster analysis. The algorithm is introduced and its performance is compared to that of the random, k-means center selection procedures and other results from the literature. By automatically defining the number of RBF centers, their positions and dispersions, the proposed method leads to parsimonious solutions. Simulation results are reported concerning regression and classification problems.     

Engineering physicochemical and biological properties of proteins by chemical modification

Proteins can be modified by chemically binding synthetic or natural macromolecules to the surface of the molecules. This can counter some of the drawbacks of the native proteins and improves properties which could be important in their use as protein drugs and/or catalysts in bioreactors. The most difficult problem is endowing delicate proteins with suitable new properties by chemical modification without causing any loss of their functions.     

The role of instability with resistance training

There are many instances in daily life and sport in which force must be exerted when an individual performing the task is in an unstable condition. Instability can decrease the externally-measured force output of a muscle while maintaining high muscle activation. The high muscle activation of limbs and trunk when unstable can be attributed to the increased stabilization functions. The increased stress associated with instability has been postulated to promote greater neuromuscular adaptations, such as decreased co-contractions, improved coordination, and confidence in performing a skill. In addition, high muscle activation with less stress on joints and muscles could also be beneficial for general musculoskeletal health and rehabilitation. However, the lower force output may be detrimental to absolute strength gains when resistance training. Furthermore, other studies have reported increased co-contractions with unstable training. The positive effects of instability resistance training on sports performance have yet to be quantified. The examination of the literature suggests that when implementing a resistance training program for musculoskeletal health or rehabilitation, both stable and unstable exercises should be included to ensure an emphasis on both higher force (stable) and balance (unstable) stressors to the neuromuscular system.     

Modelling the maximum voluntary joint torque/angular velocity relationship in human movement

The force exerted by a muscle is a function of the activation level and the maximum (tetanic) muscle force. In "maximum" voluntary knee extensions muscle activation is lower for eccentric muscle velocities than for concentric velocities. The aim of this study was to model this "differential activation" in order to calculate the maximum voluntary knee extensor torque as a function of knee angular velocity. Torque data were collected on two subjects during maximal eccentric-concentric knee extensions using an isovelocity dynamometer with crank angular velocities ranging from 50 to 450 degrees s(-1). The theoretical tetanic torque/angular velocity relationship was modelled using a four parameter function comprising two rectangular hyperbolas while the activation/angular velocity relationship was modelled using a three parameter function that rose from submaximal activation for eccentric velocities to full activation for high concentric velocities. The product of these two functions gave a seven parameter function which was fitted to the joint torque/angular velocity data, giving unbiased root mean square differences of 1.9% and 3.3% of the maximum torques achieved. Differential activation accounts for the non-hyperbolic behaviour of the torque/angular velocity data for low concentric velocities. The maximum voluntary knee extensor torque that can be exerted may be modelled accurately as the product of functions defining the maximum torque and the maximum voluntary activation level. Failure to include differential activation considerations when modelling maximal movements will lead to errors in the estimation of joint torque in the eccentric phase and low velocity concentric phase.     

A new bell-shaped function for idiotypic interactions based on cross-linking

Most recent models of the immune network are based upon a phenomenological log bell-shaped interaction function. This function depends on a single parameter, the “field”, which is the sum of all ligand concentrations weighted by their respective affinities. The typical behavior of these models is dominated by percolation, a phenomenon in which a local stimulus spreads globally throughout the network. The usual reason for employing a log bell-shaped interaction function is that B cells are activated by cross-linking of their surface immunoglobulin receptors. Here we formally derive a new phenomenological log bell-shaped function from the chemistry of receptor cross-linking by bivalent ligand. Specifying how this new function depends on the ligand concentrations requires two fields: a binding field and a cross-linking field. When we compare the activation functions for ligand-receptor pairs with different affinities, the one-field and the two-field functions differ markedly. In the case of the one-field activation function, its graph is shifted to increasingly higher concentration as the affinity decreases but keeps its width and height. In the case of the two-field activation function, the graph of a low-affinity interaction is nested within the graphs of all higher-affinity interactions. We show that this difference in the relations among activation functions for different affinities radically changes the network behavior. In models that described B cell proliferation using the one-field activation function, network behavior was dominated by low-affinity interactions. Conversely, in our new model, the high-affinity interactions are the most significant. As a consequence, percolation is no longer the only typical network behavior.