Nonspecific synaptic plasticity improves the recognition of sparse patterns degraded by local noise.

Many forms of synaptic plasticity require the local production of volatile or rapidly diffusing substances such as nitric oxide. The nonspecific plasticity these neuromodulators may induce at neighboring non-active synapses is thought to be detrimental for the specificity of memory storage. We show here that memory retrieval may benefit from this non-specific plasticity when the applied sparse binary input patterns are degraded by local noise. Simulations of a biophysically realistic model of a cerebellar Purkinje cell in a pattern recognition task show that, in the absence of noise, leakage of plasticity to adjacent synapses degrades the recognition of sparse static patterns. However, above a local noise level of 20%, the model with nonspecific plasticity outperforms the standard, specific model. The gain in performance is greatest when the spatial distribution of noise in the input matches the range of diffusion-induced plasticity. Hence non-specific plasticity may offer a benefit in noisy environments or when the pressure to generalize is strong.

Support vector regression to estimate the permeability enhancement of potential transdermal enhancers.

OBJECTIVES: Searching for chemicals that will safely enhance transdermal drug delivery is a significant challenge. This study applies support vector regression (SVR) for the first time to estimating the optimal formulation design of transdermal hydrocortisone formulations. METHODS: The aim of this study was to apply SVR methods with two different kernels in order to estimate the enhancement ratio of chemical enhancers of permeability. KEY FINDINGS: A statistically significant regression SVR model was developed. It was found that SVR with a nonlinear kernel provided the best estimate of the enhancement ratio for a chemical enhancer. CONCLUSIONS: Support vector regression is a viable method to develop predictive models of biological processes, demonstrating improvements over other methods. In addition, the results of this study suggest that a global approach to modelling a biological process may not necessarily be the best method and that a 'mixed-methods' approach may be best in optimising predictive models.

Dendritic morphology predicts pattern recognition performance in multi-compartmental model neurons with and without active conductances.

In this paper we examine how a neuron's dendritic morphology can affect its pattern recognition performance. We use two different algorithms to systematically explore the space of dendritic morphologies: an algorithm that generates all possible dendritic trees with 22 terminal points, and one that creates representative samples of trees with 128 terminal points. Based on these trees, we construct multi-compartmental models. To assess the performance of the resulting neuronal models, we quantify their ability to discriminate learnt and novel input patterns. We find that the dendritic morphology does have a considerable effect on pattern recognition performance and that the neuronal performance is inversely correlated with the mean depth of the dendritic tree. The results also reveal that the asymmetry index of the dendritic tree does not correlate with the performance for the full range of tree morphologies. The performance of neurons with dendritic tapering is best predicted by the mean and variance of the electrotonic distance of their synapses to the soma. All relationships found for passive neuron models also hold, even in more accentuated form, for neurons with active membranes.

The application and limitations of mathematical modelling in the prediction of permeability across mammalian skin and polydimethylsiloxane membranes.

OBJECTIVES: Predicting the rate of percutaneous absorption of a drug is an important issue with the increasing use of the skin as a means of moderating and controlling drug delivery. One key feature of this problem domain is that human skin permeability (as K(p)) has been shown to be inherently non-linear when mathematically related to the physicochemical parameters of penetrants. As such, the aims of this study were to apply and evaluate Gaussian process (GP) regression methods to datasets for membranes other than human skin, and to explore how the nature of the dataset may influence its analysis. METHODS: Permeability data for absorption across rodent and pig skin, and artificial membranes (polydimethylsiloxane, PDMS, i.e. Silastic) membranes was collected from the literature. Two quantitative structure-permeability relationship (QSPR) models were used to compare with the GP models. Further performance metrics were computed in terms of all predictions, and a range of covariance functions were examined: the squared exponential (SE), neural network (NNone) and rational quadratic (QR) covariance functions, along with two simple cases of Matern covariance function (Matern3 and Matern5) where the polynomial order is set to 1 and 2, respectively. As measures of performance, the correlation coefficient (CORR), negative log estimated predictive density (NLL, or negative log loss) and mean squared error (MSE) were employed. KEY FINDINGS: The results demonstrated that GP models with different covariance functions outperform QSPR models for human, pig and rodent datasets. For the artificial membranes, GPs perform better in one instance, and give similar results in other experiments (where different covariance parameters produce similar results). In some cases, the GP predictions for some of the artificial membrane dataset are poorly correlated, suggesting that the physicochemical parameters employed in this study might not be appropriate for developing models that represent this membrane. CONCLUSIONS: While the results of this study indicate that permeation across rodent (mouse and rat) and pig skin is, in a statistical sense, similar, and that the artificial membranes are poor replacements of human or animal skin, the overriding issue raised in this study is the nature of the dataset and how it can influence the results, and subsequent interpretation, of any model produced for particular membranes. The size of the datasets, in both absolute and comparative senses, appears to influence model quality. Ideally, to generate viable cross-comparisons the datasets for different mammalian membranes should, wherever possible, exhibit as much commonality as possible.

STD-dependent and independent encoding of input irregularity as spike rate in a computational model of a cerebellar nucleus neuron.

Neurons in the cerebellar nuclei (CN) receive inhibitory inputs from Purkinje cells in the cerebellar cortex and provide the major output from the cerebellum, but their computational function is not well understood. It has recently been shown that the spike activity of Purkinje cells is more regular than previously assumed and that this regularity can affect motor behaviour. We use a conductance-based model of a CN neuron to study the effect of the regularity of Purkinje cell spiking on CN neuron activity. We find that increasing the irregularity of Purkinje cell activity accelerates the CN neuron spike rate and that the mechanism of this recoding of input irregularity as output spike rate depends on the number of Purkinje cells converging onto a CN neuron. For high convergence ratios, the irregularity induced spike rate acceleration depends on short-term depression (STD) at the Purkinje cell synapses. At low convergence ratios, or for synchronised Purkinje cell input, the firing rate increase is independent of STD. The transformation of input irregularity into output spike rate occurs in response to artificial input spike trains as well as to spike trains recorded from Purkinje cells in tottering mice, which show highly irregular spiking patterns. Our results suggest that STD may contribute to the accelerated CN spike rate in tottering mice and they raise the possibility that the deficits in motor control in these mutants partly result as a pathological consequence of this natural form of plasticity.

Clustering predicts memory performance in networks of spiking and non-spiking neurons.

The problem we address in this paper is that of finding effective and parsimonious patterns of connectivity in sparse associative memories. This problem must be addressed in real neuronal systems, so that results in artificial systems could throw light on real systems. We show that there are efficient patterns of connectivity and that these patterns are effective in models with either spiking or non-spiking neurons. This suggests that there may be some underlying general principles governing good connectivity in such networks. We also show that the clustering of the network, measured by Clustering Coefficient, has a strong negative linear correlation to the performance of associative memory. This result is important since a purely static measure of network connectivity appears to determine an important dynamic property of the network.

The application of feature selection to the development of Gaussian process models for percutaneous absorption.

OBJECTIVES: The aim was to employ Gaussian processes to assess mathematically the nature of a skin permeability dataset and to employ these methods, particularly feature selection, to determine the key physicochemical descriptors which exert the most significant influence on percutaneous absorption, and to compare such models with established existing models. METHODS: Gaussian processes, including automatic relevance detection (GPRARD) methods, were employed to develop models of percutaneous absorption that identified key physicochemical descriptors of percutaneous absorption. Using MatLab software, the statistical performance of these models was compared with single linear networks (SLN) and quantitative structure-permeability relationships (QSPRs). Feature selection methods were used to examine in more detail the physicochemical parameters used in this study. A range of statistical measures to determine model quality were used. KEY FINDINGS: The inherently nonlinear nature of the skin data set was confirmed. The Gaussian process regression (GPR) methods yielded predictive models that offered statistically significant improvements over SLN and QSPR models with regard to predictivity (where the rank order was: GPR > SLN > QSPR). Feature selection analysis determined that the best GPR models were those that contained log P, melting point and the number of hydrogen bond donor groups as significant descriptors. Further statistical analysis also found that great synergy existed between certain parameters. It suggested that a number of the descriptors employed were effectively interchangeable, thus questioning the use of models where discrete variables are output, usually in the form of an equation. CONCLUSIONS: The use of a nonlinear GPR method produced models with significantly improved predictivity, compared with SLN or QSPR models. Feature selection methods were able to provide important mechanistic information. However, it was also shown that significant synergy existed between certain parameters, and as such it was possible to interchange certain descriptors (i.e. molecular weight and melting point) without incurring a loss of model quality. Such synergy suggested that a model constructed from discrete terms in an equation may not be the most appropriate way of representing mechanistic understandings of skin absorption.

The application of Gaussian processes in the prediction of percutaneous absorption.

OBJECTIVES: The aim was to assess mathematically the nature of a skin permeability dataset and to determine the utility of Gaussian processes in developing a predictive model for skin permeability, comparing it with existing methods for deriving predictive models. METHODS: Principal component analysis was carried out in order to determine the nature of the dataset. MatLab software was used to assess the performance of Gaussian process, single linear networks (SLN) and quantitative structure-permeability relationships (QSPRs) using a range of statistical measures. KEY FINDINGS: Principal component analysis showed that the dataset is inherently non-linear. The Gaussian process model yielded a predictive model that provides a significantly more accurate estimate of skin absorption than previous models, particularly QSPRs (which were consistently worse than Gaussian process or SLN models), and does so across a wider range of molecular properties. Gaussian process models appear particularly capable of providing excellent predictions where previous studies have shown QSPRs to fail, such as where penetrants have high log P and high molecular weight. CONCLUSIONS: A non-linear approach was more appropriate than QSPRs or SLNs for the analysis of the dataset employed herein, as the prediction and confidence values in the prediction given by the Gaussian process are better than with other methods examined. Gaussian process provides a novel way of analysing skin absorption data that is substantially more accurate, statistically robust and reflective of our empirical understanding of skin absorption than the QSPR methods so far applied to skin absorption.

Combining experts in order to identify binding sites in yeast and mouse genomic data.

The identification of cis-regulatory binding sites in DNA is a difficult problem in computational biology. To obtain a full understanding of the complex machinery embodied in genetic regulatory networks it is necessary to know both the identity of the regulatory transcription factors and the location of their binding sites in the genome. We show that using an SVM together with data sampling to classify the combination of the results of individual algorithms specialised for the prediction of binding site locations, can produce significant improvements upon the original algorithms. The resulting classifier produces fewer false positive predictions and so reduces the expensive experimental procedure of verifying the predictions.

Efficient connection strategies in 1D and 2D associative memory models with and without displaced connectivity.

This study examines the performance of sparsely connected associative memory models built using a number of different connection strategies, applied to one- and two-dimensional topologies. Efficient patterns of connectivity are identified which yield high performance at relatively low wiring costs in both topologies. Networks with displaced connectivity are seen to perform particularly well. It is found that two-dimensional models are more tolerant of variations in connection strategy than their one-dimensional counterparts; though networks built with both topologies become less so as their connection density is decreased.

Improving computational predictions of cis-regulatory binding sites.

The location of cis-regulatory binding sites determine the connectivity of genetic regulatory networks and therefore constitute a natural focal point for research into the many biological systems controlled by such regulatory networks. Accurate computational prediction of these binding sites would facilitate research into a multitude of key areas, including embryonic development, evolution, pharmacogenemics, cancer and many other transcriptional diseases, and is likely to be an important precursor for the reverse engineering of genome wide, genetic regulatory networks. Many algorithmic strategies have been developed for the computational prediction of cis-regulatory binding sites but currently all approaches are prone to high rates of false positive predictions, and many are highly dependent on additional information, limiting their usefulness as research tools. In this paper we present an approach for improving the accuracy of a selection of established prediction algorithms. Firstly, it is shown that species specific optimization of algorithmic parameters can, in some cases, significantly improve the accuracy of algorithmic predictions. Secondly, it is demonstrated that the use of non-linear classification algorithms to integrate predictions from multiple sources can result in more accurate predictions. Finally, it is shown that further improvements in prediction accuracy can be gained with the use of biologically inspired post-processing of predictions.