Evaluating different methods of microarray data normalization

Abstract Background With the development of DNA hybridization microarray technologies, nowadays it is possible to simultaneously assess the expression levels of thousands to tens of thousands of genes. Quantitative comparison of microarrays uncovers distinct patterns of gene expression, which define different cellular phenotypes or cellular responses to drugs. Due to technical biases, normalization of the intensity levels is a pre-requisite to performing further statistical analyses. Therefore, choosing a suitable approach for normalization can be critical, deserving judicious consideration. Results Here, we considered three commonly used normalization approaches, namely: Loess, Splines and Wavelets, and two non-parametric regression methods, which have yet to be used for normalization, namely, the Kernel smoothing and Support Vector Regression. The results obtained were compared using artificial microarray data and benchmark studies. The results indicate that the Support Vector Regression is the most robust to outliers and that Kernel is the worst normalization technique, while no practical differences were observed between Loess, Splines and Wavelets. Conclusion In face of our results, the Support Vector Regression is favored for microarray normalization due to its superiority when compared to the other methods for its robustness in estimating the normalization curve.

Procalcitonin (PCT) and C-reactive Protein (CRP) as severe systemic infection markers in febrile neutropenic adults

Abstract Background Procalcitonin (PCT) is an inflammatory marker that has been used as indicator of severe bacterial infection. We evaluated the concentrations of PCT as a marker for systemic infection compared to C-reactive protein (CRP) in patients neutropenic febrile. Methods 52 adult patients were enrolled in the study. Blood sample was collected in order to determine the serum concentrations of PCT, CRP and other hematological parameters at the onset of fever. The patients were divided into 2 groups, one with severe infection (n = 26) and the other in which the patients did not present such an infection (n = 26). Then PCT and CRP concentrations at the fever onset were compared between groups using non parametric statistical tests, ROC curve, sensitivity, specificity, likelihood ratio, and Spearman's correlation coefficient. Results The mean of PCT was significantly higher in the group with severe infection (6.7 ng/mL versus 0.6 ng/mL – p = 0.0075) comparing with CRP. Serum concentrations of 0.245 ng/mL of PCT displayed 100% de sensitivity and 69.2% specificity. PCT concentrations of 2,145 ng/mL presented a likelihood ratio of 13, which was not observed for any concentration of CRP. Conclusion PCT seems to be an useful marker for the diagnosis of systemic infection in febrile neutropenic patients, probably better than CRP.

A partial dependence factorial analysis to deal with selection bias in observational studis

This thesis presents a creative and practical approach to dealing with the problem of selection bias. Selection bias may be the most important vexing problem in program evaluation or in any line of research that attempts to assert causality. Some of the greatest minds in economics and statistics have scrutinized the problem of selection bias, with the resulting approaches – Rubin’s Potential Outcome Approach(Rosenbaum and Rubin,1983; Rubin, 1991,2001,2004) or Heckman’s Selection model (Heckman, 1979) – being widely accepted and used as the best fixes. These solutions to the bias that arises in particular from self selection are imperfect, and many researchers, when feasible, reserve their strongest causal inference for data from experimental rather than observational studies. The innovative aspect of this thesis is to propose a data transformation that allows measuring and testing in an automatic and multivariate way the presence of selection bias. The approach involves the construction of a multi-dimensional conditional space of the X matrix in which the bias associated with the treatment assignment has been eliminated. Specifically, we propose the use of a partial dependence analysis of the X-space as a tool for investigating the dependence relationship between a set of observable pre-treatment categorical covariates X and a treatment indicator variable T, in order to obtain a measure of bias according to their dependence structure. The measure of selection bias is then expressed in terms of inertia due to the dependence between X and T that has been eliminated. Given the measure of selection bias, we propose a multivariate test of imbalance in order to check if the detected bias is significant, by using the asymptotical distribution of inertia due to T (Estadella et al. 2005) , and by preserving the multivariate nature of data. Further, we propose the use of a clustering procedure as a tool to find groups of comparable units on which estimate local causal effects, and the use of the multivariate test of imbalance as a stopping rule in choosing the best cluster solution set. The method is non parametric, it does not call for modeling the data, based on some underlying theory or assumption about the selection process, but instead it calls for using the existing variability within the data and letting the data to speak. The idea of proposing this multivariate approach to measure selection bias and test balance comes from the consideration that in applied research all aspects of multivariate balance, not represented in the univariate variable- by-variable summaries, are ignored. The first part contains an introduction to evaluation methods as part of public and private decision process and a review of the literature of evaluation methods. The attention is focused on Rubin Potential Outcome Approach, matching methods, and briefly on Heckman’s Selection Model. The second part focuses on some resulting limitations of conventional methods, with particular attention to the problem of how testing in the correct way balancing. The third part contains the original contribution proposed , a simulation study that allows to check the performance of the method for a given dependence setting and an application to a real data set. Finally, we discuss, conclude and explain our future perspectives.

A randomized, controlled clinical trial on the clinical, microbiological, and staining effects of a novel 0.05% chlorhexidine/herbal extract and a 0.1% chlorhexidine mouthrinse adjunct to periodontal surgery

BACKGROUND: Chlorhexidine (CHX) rinsing after periodontal surgery is common. We assessed the clinical and microbiological effects of two CHX concentrations following periodontal surgery. MATERIALS AND METHODS: In a randomized, controlled clinical trial, 45 subjects were assigned to 4 weeks rinsing with a 0.05 CHX/herbal extract combination (test) or a 0.1% CHX solution. Clinical and staining effects were studied. Subgingival bacteria were assessed using the DNA-DNA checkerboard. Statistics included parametric and non-parametric tests (p<0001 to declare significance at 80% power). RESULTS: At weeks 4 and 12, more staining was found in the control group (p<0.05 and p<0.001, respectively). A higher risk for staining was found in the control group (crude OR: 2.3:1, 95% CI: 1.3 to 4.4, p<0.01). The absolute staining reduction in the test group was 21.1% (9 5% CI: 9.4-32.8%). Probing pocket depth (PPD) decreases were significant (p<0.001) in both groups and similar (p=0.92). No rinse group differences in changes of bacterial counts for any species were found between baseline and week 12. CONCLUSIONS: The test CHX rinse resulted in less tooth staining. At the study endpoint, similar and high counts of periodontal pathogens were found.

Development of a model-based transient calibration process for diesel engine electronic control module tables – Part 1: data requirements, processing, and analysis

This is the first part of a study investigating a model-based transient calibration process for diesel engines. The motivation is to populate hundreds of parameters (which can be calibrated) in a methodical and optimum manner by using model-based optimization in conjunction with the manual process so that, relative to the manual process used by itself, a significant improvement in transient emissions and fuel consumption and a sizable reduction in calibration time and test cell requirements is achieved. Empirical transient modelling and optimization has been addressed in the second part of this work, while the required data for model training and generalization are the focus of the current work. Transient and steady-state data from a turbocharged multicylinder diesel engine have been examined from a model training perspective. A single-cylinder engine with external air-handling has been used to expand the steady-state data to encompass transient parameter space. Based on comparative model performance and differences in the non-parametric space, primarily driven by a high engine difference between exhaust and intake manifold pressures (ΔP) during transients, it has been recommended that transient emission models should be trained with transient training data. It has been shown that electronic control module (ECM) estimates of transient charge flow and the exhaust gas recirculation (EGR) fraction cannot be accurate at the high engine ΔP frequently encountered during transient operation, and that such estimates do not account for cylinder-to-cylinder variation. The effects of high engine ΔP must therefore be incorporated empirically by using transient data generated from a spectrum of transient calibrations. Specific recommendations on how to choose such calibrations, how many data to acquire, and how to specify transient segments for data acquisition have been made. Methods to process transient data to account for transport delays and sensor lags have been developed. The processed data have then been visualized using statistical means to understand transient emission formation. Two modes of transient opacity formation have been observed and described. The first mode is driven by high engine ΔP and low fresh air flowrates, while the second mode is driven by high engine ΔP and high EGR flowrates. The EGR fraction is inaccurately estimated at both modes, while EGR distribution has been shown to be present but unaccounted for by the ECM. The two modes and associated phenomena are essential to understanding why transient emission models are calibration dependent and furthermore how to choose training data that will result in good model generalization.

A Decision-Tree-Based Approach to Smoke Spike Detection in a Heavy-Duty Diesel Engine

Smoke spikes occurring during transient engine operation have detrimental health effects and increase fuel consumption by requiring more frequent regeneration of the diesel particulate filter. This paper proposes a decision tree approach to real-time detection of smoke spikes for control and on-board diagnostics purposes. A contemporary, electronically controlled heavy-duty diesel engine was used to investigate the deficiencies of smoke control based on the fuel-to-oxygen-ratio limit. With the aid of transient and steady state data analysis and empirical as well as dimensional modeling, it was shown that the fuel-to-oxygen ratio was not estimated correctly during the turbocharger lag period. This inaccuracy was attributed to the large manifold pressure ratios and low exhaust gas recirculation flows recorded during the turbocharger lag period, which meant that engine control module correlations for the exhaust gas recirculation flow and the volumetric efficiency had to be extrapolated. The engine control module correlations were based on steady state data and it was shown that, unless the turbocharger efficiency is artificially reduced, the large manifold pressure ratios observed during the turbocharger lag period cannot be achieved at steady state. Additionally, the cylinder-to-cylinder variation during this period were shown to be sufficiently significant to make the average fuel-to-oxygen ratio a poor predictor of the transient smoke emissions. The steady state data also showed higher smoke emissions with higher exhaust gas recirculation fractions at constant fuel-to-oxygen-ratio levels. This suggests that, even if the fuel-to-oxygen ratios were to be estimated accurately for each cylinder, they would still be ineffective as smoke limiters. A decision tree trained on snap throttle data and pruned with engineering knowledge was able to use the inaccurate engine control module estimates of the fuel-to-oxygen ratio together with information on the engine control module estimate of the exhaust gas recirculation fraction, the engine speed, and the manifold pressure ratio to predict 94% of all spikes occurring over the Federal Test Procedure cycle. The advantages of this non-parametric approach over other commonly used parametric empirical methods such as regression were described. An application of accurate smoke spike detection in which the injection pressure is increased at points with a high opacity to reduce the cumulative particulate matter emissions substantially with a minimum increase in the cumulative nitrogrn oxide emissions was illustrated with dimensional and empirical modeling.

Advantages and Applications of Transforming Empirical Model Input Space with Dimensional Models

Model-based calibration of steady-state engine operation is commonly performed with highly parameterized empirical models that are accurate but not very robust, particularly when predicting highly nonlinear responses such as diesel smoke emissions. To address this problem, and to boost the accuracy of more robust non-parametric methods to the same level, GT-Power was used to transform the empirical model input space into multiple input spaces that simplified the input-output relationship and improved the accuracy and robustness of smoke predictions made by three commonly used empirical modeling methods: Multivariate Regression, Neural Networks and the k-Nearest Neighbor method. The availability of multiple input spaces allowed the development of two committee techniques: a 'Simple Committee' technique that used averaged predictions from a set of 10 pre-selected input spaces chosen by the training data and the "Minimum Variance Committee" technique where the input spaces for each prediction were chosen on the basis of disagreement between the three modeling methods. This latter technique equalized the performance of the three modeling methods. The successively increasing improvements resulting from the use of a single best transformed input space (Best Combination Technique), Simple Committee Technique and Minimum Variance Committee Technique were verified with hypothesis testing. The transformed input spaces were also shown to improve outlier detection and to improve k-Nearest Neighbor performance when predicting dynamic emissions with steady-state training data. An unexpected finding was that the benefits of input space transformation were unaffected by changes in the hardware or the calibration of the underlying GT-Power model.

Truncated recombinant human SP-D attenuates emphysema and type II cell changes in SP-D deficient mice

BACKGROUND: Surfactant protein D (SP-D) deficient mice develop emphysema-like pathology associated with focal accumulations of foamy alveolar macrophages, an excess of surfactant phospholipids in the alveolar space and both hypertrophy and hyperplasia of alveolar type II cells. These findings are associated with a chronic inflammatory state. Treatment of SP-D deficient mice with a truncated recombinant fragment of human SP-D (rfhSP-D) has been shown to decrease the lipidosis and alveolar macrophage accumulation as well as production of proinflammatory chemokines. The aim of this study was to investigate if rfhSP-D treatment reduces the structural abnormalities in parenchymal architecture and type II cells characteristic of SP-D deficiency. METHODS: SP-D knock-out mice, aged 3 weeks, 6 weeks and 9 weeks were treated with rfhSP-D for 9, 6 and 3 weeks, respectively. All mice were sacrificed at age 12 weeks and compared to both PBS treated SP-D deficient and wild-type groups. Lung structure was quantified by design-based stereology at the light and electron microscopic level. Emphasis was put on quantification of emphysema, type II cell changes and intracellular surfactant. Data were analysed with two sided non-parametric Mann-Whitney U-test. MAIN RESULTS: After 3 weeks of treatment, alveolar number was higher and mean alveolar size was smaller compared to saline-treated SP-D knock-out controls. There was no significant difference concerning these indices of pulmonary emphysema within rfhSP-D treated groups. Type II cell number and size were smaller as a consequence of treatment. The total volume of lamellar bodies per type II cell and per lung was smaller after 6 weeks of treatment. CONCLUSION: Treatment of SP-D deficient mice with rfhSP-D leads to a reduction in the degree of emphysema and a correction of type II cell hyperplasia and hypertrophy. This supports the concept that rfhSP-D might become a therapeutic option in diseases that are characterized by decreased SP-D levels in the lung.

Comparison of two non-bronchoscopic methods for evaluating inflammation in patients with acute hypoxaemic respiratory failure

INTRODUCTION: The simple bedside method for sampling undiluted distal pulmonary edema fluid through a normal suction catheter (s-Cath) has been experimentally and clinically validated. However, there are no data comparing non-bronchoscopic bronchoalveolar lavage (mini-BAL) and s-Cath for assessing lung inflammation in acute hypoxaemic respiratory failure. We designed a prospective study in two groups of patients, those with acute lung injury (ALI)/acute respiratory distress syndrome (ARDS) and those with acute cardiogenic lung edema (ACLE), designed to investigate the clinical feasibility of these techniques and to evaluate inflammation in both groups using undiluted sampling obtained by s-Cath. To test the interchangeability of the two methods in the same patient for studying the inflammation response, we further compared mini-BAL and s-Cath for agreement of protein concentration and percentage of polymorphonuclear cells (PMNs). METHODS: Mini-BAL and s-Cath sampling was assessed in 30 mechanically ventilated patients, 21 with ALI/ARDS and 9 with ACLE. To analyse agreement between the two sampling techniques, we considered only simultaneously collected mini-BAL and s-Cath paired samples. The protein concentration and polymorphonuclear cell (PMN) count comparisons were performed using undiluted sampling. Bland-Altman plots were used for assessing the mean bias and the limits of agreement between the two sampling techniques; comparison between groups was performed by using the non-parametric Mann-Whitney-U test; continuous variables were compared by using the Student t-test, Wilcoxon signed rank test, analysis of variance or Student-Newman-Keuls test; and categorical variables were compared by using chi-square analysis or Fisher exact test. RESULTS: Using protein content and PMN percentage as parameters, we identified substantial variations between the two sampling techniques. When the protein concentration in the lung was high, the s-Cath was a more sensitive method; by contrast, as inflammation increased, both methods provided similar estimates of neutrophil percentages in the lung. The patients with ACLE showed an increased PMN count, suggesting that hydrostatic lung edema can be associated with a concomitant inflammatory process. CONCLUSIONS: There are significant differences between the s-Cath and mini-BAL sampling techniques, indicating that these procedures cannot be used interchangeably for studying the lung inflammatory response in patients with acute hypoxaemic lung injury.

A BAYESIAN APPROACH TO DOSE-RESPONSE ASSESSMENT AND DRUG-DRUG INTERACTION ANALYSIS: APPLICATION TO IN VITRO STUDIES

The considerable search for synergistic agents in cancer research is motivated by the therapeutic benefits achieved by combining anti-cancer agents. Synergistic agents make it possible to reduce dosage while maintaining or enhancing a desired effect. Other favorable outcomes of synergistic agents include reduction in toxicity and minimizing or delaying drug resistance. Dose-response assessment and drug-drug interaction analysis play an important part in the drug discovery process, however analysis are often poorly done. This dissertation is an effort to notably improve dose-response assessment and drug-drug interaction analysis. The most commonly used method in published analysis is the Median-Effect Principle/Combination Index method (Chou and Talalay, 1984). The Median-Effect Principle/Combination Index method leads to inefficiency by ignoring important sources of variation inherent in dose-response data and discarding data points that do not fit the Median-Effect Principle. Previous work has shown that the conventional method yields a high rate of false positives (Boik, Boik, Newman, 2008; Hennessey, Rosner, Bast, Chen, 2010) and, in some cases, low power to detect synergy. There is a great need for improving the current methodology. We developed a Bayesian framework for dose-response modeling and drug-drug interaction analysis. First, we developed a hierarchical meta-regression dose-response model that accounts for various sources of variation and uncertainty and allows one to incorporate knowledge from prior studies into the current analysis, thus offering a more efficient and reliable inference. Second, in the case that parametric dose-response models do not fit the data, we developed a practical and flexible nonparametric regression method for meta-analysis of independently repeated dose-response experiments. Third, and lastly, we developed a method, based on Loewe additivity that allows one to quantitatively assess interaction between two agents combined at a fixed dose ratio. The proposed method makes a comprehensive and honest account of uncertainty within drug interaction assessment. Extensive simulation studies show that the novel methodology improves the screening process of effective/synergistic agents and reduces the incidence of type I error. We consider an ovarian cancer cell line study that investigates the combined effect of DNA methylation inhibitors and histone deacetylation inhibitors in human ovarian cancer cell lines. The hypothesis is that the combination of DNA methylation inhibitors and histone deacetylation inhibitors will enhance antiproliferative activity in human ovarian cancer cell lines compared to treatment with each inhibitor alone. By applying the proposed Bayesian methodology, in vitro synergy was declared for DNA methylation inhibitor, 5-AZA-2'-deoxycytidine combined with one histone deacetylation inhibitor, suberoylanilide hydroxamic acid or trichostatin A in the cell lines HEY and SKOV3. This suggests potential new epigenetic therapies in cell growth inhibition of ovarian cancer cells.

Street crossing behavior in younger and older pedestrians: an eye- and head-tracking study

BACKGROUND: Crossing a street can be a very difficult task for older pedestrians. With increased age and potential cognitive decline, older people take the decision to cross a street primarily based on vehicles' distance, and not on their speed. Furthermore, older pedestrians tend to overestimate their own walking speed, and could not adapt it according to the traffic conditions. Pedestrians' behavior is often tested using virtual reality. Virtual reality presents the advantage of being safe, cost-effective, and allows using standardized test conditions. METHODS: This paper describes an observational study with older and younger adults. Street crossing behavior was investigated in 18 healthy, younger and 18 older subjects by using a virtual reality setting. The aim of the study was to measure behavioral data (such as eye and head movements) and to assess how the two age groups differ in terms of number of safe street crossings, virtual crashes, and missed street crossing opportunities. Street crossing behavior, eye and head movements, in older and younger subjects, were compared with non-parametric tests. RESULTS: The results showed that younger pedestrians behaved in a more secure manner while crossing a street, as compared to older people. The eye and head movements analysis revealed that older people looked more at the ground and less at the other side of the street to cross. CONCLUSIONS: The less secure behavior in street crossing found in older pedestrians could be explained by their reduced cognitive and visual abilities, which, in turn, resulted in difficulties in the decision-making process, especially under time pressure. Decisions to cross a street are based on the distance of the oncoming cars, rather than their speed, for both groups. Older pedestrians look more at their feet, probably because of their need of more time to plan precise stepping movement and, in turn, pay less attention to the traffic. This might help to set up guidelines for improving senior pedestrians' safety, in terms of speed limits, road design, and mixed physical-cognitive trainings.

A randomized trial of the effects of the noble gases helium and argon on neuroprotection in a rodent cardiac arrest model.

BACKGROUND The noble gas xenon is considered as a neuroprotective agent, but availability of the gas is limited. Studies on neuroprotection with the abundant noble gases helium and argon demonstrated mixed results, and data regarding neuroprotection after cardiac arrest are scant. We tested the hypothesis that administration of 50% helium or 50% argon for 24 h after resuscitation from cardiac arrest improves clinical and histological outcome in our 8 min rat cardiac arrest model. METHODS Forty animals had cardiac arrest induced with intravenous potassium/esmolol and were randomized to post-resuscitation ventilation with either helium/oxygen, argon/oxygen or air/oxygen for 24 h. Eight additional animals without cardiac arrest served as reference, these animals were not randomized and not included into the statistical analysis. Primary outcome was assessment of neuronal damage in histology of the region I of hippocampus proper (CA1) from those animals surviving until day 5. Secondary outcome was evaluation of neurobehavior by daily testing of a Neurodeficit Score (NDS), the Tape Removal Test (TRT), a simple vertical pole test (VPT) and the Open Field Test (OFT). Because of the non-parametric distribution of the data, the histological assessments were compared with the Kruskal-Wallis test. Treatment effect in repeated measured assessments was estimated with a linear regression with clustered robust standard errors (SE), where normality is less important. RESULTS Twenty-nine out of 40 rats survived until day 5 with significant initial deficits in neurobehavioral, but rapid improvement within all groups randomized to cardiac arrest. There were no statistical significant differences between groups neither in the histological nor in neurobehavioral assessment. CONCLUSIONS The replacement of air with either helium or argon in a 50:50 air/oxygen mixture for 24 h did not improve histological or clinical outcome in rats subjected to 8 min of cardiac arrest.

Comparative analysis of meta-analysis methods: when to use which?

Background: Several meta-analysis methods can be used to quantitatively combine the results of a group of experiments, including the weighted mean difference, statistical vote counting, the parametric response ratio and the non-parametric response ratio. The software engineering community has focused on the weighted mean difference method. However, other meta-analysis methods have distinct strengths, such as being able to be used when variances are not reported. There are as yet no guidelines to indicate which method is best for use in each case. Aim: Compile a set of rules that SE researchers can use to ascertain which aggregation method is best for use in the synthesis phase of a systematic review. Method: Monte Carlo simulation varying the number of experiments in the meta analyses, the number of subjects that they include, their variance and effect size. We empirically calculated the reliability and statistical power in each case Results: WMD is generally reliable if the variance is low, whereas its power depends on the effect size and number of subjects per meta-analysis; the reliability of RR is generally unaffected by changes in variance, but it does require more subjects than WMD to be powerful; NPRR is the most reliable method, but it is not very powerful; SVC behaves well when the effect size is moderate, but is less reliable with other effect sizes. Detailed tables of results are annexed. Conclusions: Before undertaking statistical aggregation in software engineering, it is worthwhile checking whether there is any appreciable difference in the reliability and power of the methods. If there is, software engineers should select the method that optimizes both parameters.

Regularization for sparsity in statistical analysis and machine learning

Pragmatism is the leading motivation of regularization. We can understand regularization as a modification of the maximum-likelihood estimator so that a reasonable answer could be given in an unstable or ill-posed situation. To mention some typical examples, this happens when fitting parametric or non-parametric models with more parameters than data or when estimating large covariance matrices. Regularization is usually used, in addition, to improve the bias-variance tradeoff of an estimation. Then, the definition of regularization is quite general, and, although the introduction of a penalty is probably the most popular type, it is just one out of multiple forms of regularization. In this dissertation, we focus on the applications of regularization for obtaining sparse or parsimonious representations, where only a subset of the inputs is used. A particular form of regularization, L1-regularization, plays a key role for reaching sparsity. Most of the contributions presented here revolve around L1-regularization, although other forms of regularization are explored (also pursuing sparsity in some sense). In addition to present a compact review of L1-regularization and its applications in statistical and machine learning, we devise methodology for regression, supervised classification and structure induction of graphical models. Within the regression paradigm, we focus on kernel smoothing learning, proposing techniques for kernel design that are suitable for high dimensional settings and sparse regression functions. We also present an application of regularized regression techniques for modeling the response of biological neurons. Supervised classification advances deal, on the one hand, with the application of regularization for obtaining a na¨ıve Bayes classifier and, on the other hand, with a novel algorithm for brain-computer interface design that uses group regularization in an efficient manner. Finally, we present a heuristic for inducing structures of Gaussian Bayesian networks using L1-regularization as a filter. El pragmatismo es la principal motivación de la regularización. Podemos entender la regularización como una modificación del estimador de máxima verosimilitud, de tal manera que se pueda dar una respuesta cuando la configuración del problema es inestable. A modo de ejemplo, podemos mencionar el ajuste de modelos paramétricos o no paramétricos cuando hay más parámetros que casos en el conjunto de datos, o la estimación de grandes matrices de covarianzas. Se suele recurrir a la regularización, además, para mejorar el compromiso sesgo-varianza en una estimación. Por tanto, la definición de regularización es muy general y, aunque la introducción de una función de penalización es probablemente el método más popular, éste es sólo uno de entre varias posibilidades. En esta tesis se ha trabajado en aplicaciones de regularización para obtener representaciones dispersas, donde sólo se usa un subconjunto de las entradas. En particular, la regularización L1 juega un papel clave en la búsqueda de dicha dispersión. La mayor parte de las contribuciones presentadas en la tesis giran alrededor de la regularización L1, aunque también se exploran otras formas de regularización (que igualmente persiguen un modelo disperso). Además de presentar una revisión de la regularización L1 y sus aplicaciones en estadística y aprendizaje de máquina, se ha desarrollado metodología para regresión, clasificación supervisada y aprendizaje de estructura en modelos gráficos. Dentro de la regresión, se ha trabajado principalmente en métodos de regresión local, proponiendo técnicas de diseño del kernel que sean adecuadas a configuraciones de alta dimensionalidad y funciones de regresión dispersas. También se presenta una aplicación de las técnicas de regresión regularizada para modelar la respuesta de neuronas reales. Los avances en clasificación supervisada tratan, por una parte, con el uso de regularización para obtener un clasificador naive Bayes y, por otra parte, con el desarrollo de un algoritmo que usa regularización por grupos de una manera eficiente y que se ha aplicado al diseño de interfaces cerebromáquina. Finalmente, se presenta una heurística para inducir la estructura de redes Bayesianas Gaussianas usando regularización L1 a modo de filtro.

Contributions to Bayesian network learning with applications to neuroscience

Neuronal morphology is a key feature in the study of brain circuits, as it is highly related to information processing and functional identification. Neuronal morphology affects the process of integration of inputs from other neurons and determines the neurons which receive the output of the neurons. Different parts of the neurons can operate semi-independently according to the spatial location of the synaptic connections. As a result, there is considerable interest in the analysis of the microanatomy of nervous cells since it constitutes an excellent tool for better understanding cortical function. However, the morphologies, molecular features and electrophysiological properties of neuronal cells are extremely variable. Except for some special cases, this variability makes it hard to find a set of features that unambiguously define a neuronal type. In addition, there are distinct types of neurons in particular regions of the brain. This morphological variability makes the analysis and modeling of neuronal morphology a challenge. Uncertainty is a key feature in many complex real-world problems. Probability theory provides a framework for modeling and reasoning with uncertainty. Probabilistic graphical models combine statistical theory and graph theory to provide a tool for managing domains with uncertainty. In particular, we focus on Bayesian networks, the most commonly used probabilistic graphical model. In this dissertation, we design new methods for learning Bayesian networks and apply them to the problem of modeling and analyzing morphological data from neurons. The morphology of a neuron can be quantified using a number of measurements, e.g., the length of the dendrites and the axon, the number of bifurcations, the direction of the dendrites and the axon, etc. These measurements can be modeled as discrete or continuous data. The continuous data can be linear (e.g., the length or the width of a dendrite) or directional (e.g., the direction of the axon). These data may follow complex probability distributions and may not fit any known parametric distribution. Modeling this kind of problems using hybrid Bayesian networks with discrete, linear and directional variables poses a number of challenges regarding learning from data, inference, etc. In this dissertation, we propose a method for modeling and simulating basal dendritic trees from pyramidal neurons using Bayesian networks to capture the interactions between the variables in the problem domain. A complete set of variables is measured from the dendrites, and a learning algorithm is applied to find the structure and estimate the parameters of the probability distributions included in the Bayesian networks. Then, a simulation algorithm is used to build the virtual dendrites by sampling values from the Bayesian networks, and a thorough evaluation is performed to show the model’s ability to generate realistic dendrites. In this first approach, the variables are discretized so that discrete Bayesian networks can be learned and simulated. Then, we address the problem of learning hybrid Bayesian networks with different kinds of variables. Mixtures of polynomials have been proposed as a way of representing probability densities in hybrid Bayesian networks. We present a method for learning mixtures of polynomials approximations of one-dimensional, multidimensional and conditional probability densities from data. The method is based on basis spline interpolation, where a density is approximated as a linear combination of basis splines. The proposed algorithms are evaluated using artificial datasets. We also use the proposed methods as a non-parametric density estimation technique in Bayesian network classifiers. Next, we address the problem of including directional data in Bayesian networks. These data have some special properties that rule out the use of classical statistics. Therefore, different distributions and statistics, such as the univariate von Mises and the multivariate von Mises–Fisher distributions, should be used to deal with this kind of information. In particular, we extend the naive Bayes classifier to the case where the conditional probability distributions of the predictive variables given the class follow either of these distributions. We consider the simple scenario, where only directional predictive variables are used, and the hybrid case, where discrete, Gaussian and directional distributions are mixed. The classifier decision functions and their decision surfaces are studied at length. Artificial examples are used to illustrate the behavior of the classifiers. The proposed classifiers are empirically evaluated over real datasets. We also study the problem of interneuron classification. An extensive group of experts is asked to classify a set of neurons according to their most prominent anatomical features. A web application is developed to retrieve the experts’ classifications. We compute agreement measures to analyze the consensus between the experts when classifying the neurons. Using Bayesian networks and clustering algorithms on the resulting data, we investigate the suitability of the anatomical terms and neuron types commonly used in the literature. Additionally, we apply supervised learning approaches to automatically classify interneurons using the values of their morphological measurements. Then, a methodology for building a model which captures the opinions of all the experts is presented. First, one Bayesian network is learned for each expert, and we propose an algorithm for clustering Bayesian networks corresponding to experts with similar behaviors. Then, a Bayesian network which represents the opinions of each group of experts is induced. Finally, a consensus Bayesian multinet which models the opinions of the whole group of experts is built. A thorough analysis of the consensus model identifies different behaviors between the experts when classifying the interneurons in the experiment. A set of characterizing morphological traits for the neuronal types can be defined by performing inference in the Bayesian multinet. These findings are used to validate the model and to gain some insights into neuron morphology. Finally, we study a classification problem where the true class label of the training instances is not known. Instead, a set of class labels is available for each instance. This is inspired by the neuron classification problem, where a group of experts is asked to individually provide a class label for each instance. We propose a novel approach for learning Bayesian networks using count vectors which represent the number of experts who selected each class label for each instance. These Bayesian networks are evaluated using artificial datasets from supervised learning problems. Resumen La morfología neuronal es una característica clave en el estudio de los circuitos cerebrales, ya que está altamente relacionada con el procesado de información y con los roles funcionales. La morfología neuronal afecta al proceso de integración de las señales de entrada y determina las neuronas que reciben las salidas de otras neuronas. Las diferentes partes de la neurona pueden operar de forma semi-independiente de acuerdo a la localización espacial de las conexiones sinápticas. Por tanto, existe un interés considerable en el análisis de la microanatomía de las células nerviosas, ya que constituye una excelente herramienta para comprender mejor el funcionamiento de la corteza cerebral. Sin embargo, las propiedades morfológicas, moleculares y electrofisiológicas de las células neuronales son extremadamente variables. Excepto en algunos casos especiales, esta variabilidad morfológica dificulta la definición de un conjunto de características que distingan claramente un tipo neuronal. Además, existen diferentes tipos de neuronas en regiones particulares del cerebro. La variabilidad neuronal hace que el análisis y el modelado de la morfología neuronal sean un importante reto científico. La incertidumbre es una propiedad clave en muchos problemas reales. La teoría de la probabilidad proporciona un marco para modelar y razonar bajo incertidumbre. Los modelos gráficos probabilísticos combinan la teoría estadística y la teoría de grafos con el objetivo de proporcionar una herramienta con la que trabajar bajo incertidumbre. En particular, nos centraremos en las redes bayesianas, el modelo más utilizado dentro de los modelos gráficos probabilísticos. En esta tesis hemos diseñado nuevos métodos para aprender redes bayesianas, inspirados por y aplicados al problema del modelado y análisis de datos morfológicos de neuronas. La morfología de una neurona puede ser cuantificada usando una serie de medidas, por ejemplo, la longitud de las dendritas y el axón, el número de bifurcaciones, la dirección de las dendritas y el axón, etc. Estas medidas pueden ser modeladas como datos continuos o discretos. A su vez, los datos continuos pueden ser lineales (por ejemplo, la longitud o la anchura de una dendrita) o direccionales (por ejemplo, la dirección del axón). Estos datos pueden llegar a seguir distribuciones de probabilidad muy complejas y pueden no ajustarse a ninguna distribución paramétrica conocida. El modelado de este tipo de problemas con redes bayesianas híbridas incluyendo variables discretas, lineales y direccionales presenta una serie de retos en relación al aprendizaje a partir de datos, la inferencia, etc. En esta tesis se propone un método para modelar y simular árboles dendríticos basales de neuronas piramidales usando redes bayesianas para capturar las interacciones entre las variables del problema. Para ello, se mide un amplio conjunto de variables de las dendritas y se aplica un algoritmo de aprendizaje con el que se aprende la estructura y se estiman los parámetros de las distribuciones de probabilidad que constituyen las redes bayesianas. Después, se usa un algoritmo de simulación para construir dendritas virtuales mediante el muestreo de valores de las redes bayesianas. Finalmente, se lleva a cabo una profunda evaluaci ón para verificar la capacidad del modelo a la hora de generar dendritas realistas. En esta primera aproximación, las variables fueron discretizadas para poder aprender y muestrear las redes bayesianas. A continuación, se aborda el problema del aprendizaje de redes bayesianas con diferentes tipos de variables. Las mixturas de polinomios constituyen un método para representar densidades de probabilidad en redes bayesianas híbridas. Presentamos un método para aprender aproximaciones de densidades unidimensionales, multidimensionales y condicionales a partir de datos utilizando mixturas de polinomios. El método se basa en interpolación con splines, que aproxima una densidad como una combinación lineal de splines. Los algoritmos propuestos se evalúan utilizando bases de datos artificiales. Además, las mixturas de polinomios son utilizadas como un método no paramétrico de estimación de densidades para clasificadores basados en redes bayesianas. Después, se estudia el problema de incluir información direccional en redes bayesianas. Este tipo de datos presenta una serie de características especiales que impiden el uso de las técnicas estadísticas clásicas. Por ello, para manejar este tipo de información se deben usar estadísticos y distribuciones de probabilidad específicos, como la distribución univariante von Mises y la distribución multivariante von Mises–Fisher. En concreto, en esta tesis extendemos el clasificador naive Bayes al caso en el que las distribuciones de probabilidad condicionada de las variables predictoras dada la clase siguen alguna de estas distribuciones. Se estudia el caso base, en el que sólo se utilizan variables direccionales, y el caso híbrido, en el que variables discretas, lineales y direccionales aparecen mezcladas. También se estudian los clasificadores desde un punto de vista teórico, derivando sus funciones de decisión y las superficies de decisión asociadas. El comportamiento de los clasificadores se ilustra utilizando bases de datos artificiales. Además, los clasificadores son evaluados empíricamente utilizando bases de datos reales. También se estudia el problema de la clasificación de interneuronas. Desarrollamos una aplicación web que permite a un grupo de expertos clasificar un conjunto de neuronas de acuerdo a sus características morfológicas más destacadas. Se utilizan medidas de concordancia para analizar el consenso entre los expertos a la hora de clasificar las neuronas. Se investiga la idoneidad de los términos anatómicos y de los tipos neuronales utilizados frecuentemente en la literatura a través del análisis de redes bayesianas y la aplicación de algoritmos de clustering. Además, se aplican técnicas de aprendizaje supervisado con el objetivo de clasificar de forma automática las interneuronas a partir de sus valores morfológicos. A continuación, se presenta una metodología para construir un modelo que captura las opiniones de todos los expertos. Primero, se genera una red bayesiana para cada experto y se propone un algoritmo para agrupar las redes bayesianas que se corresponden con expertos con comportamientos similares. Después, se induce una red bayesiana que modela la opinión de cada grupo de expertos. Por último, se construye una multired bayesiana que modela las opiniones del conjunto completo de expertos. El análisis del modelo consensuado permite identificar diferentes comportamientos entre los expertos a la hora de clasificar las neuronas. Además, permite extraer un conjunto de características morfológicas relevantes para cada uno de los tipos neuronales mediante inferencia con la multired bayesiana. Estos descubrimientos se utilizan para validar el modelo y constituyen información relevante acerca de la morfología neuronal. Por último, se estudia un problema de clasificación en el que la etiqueta de clase de los datos de entrenamiento es incierta. En cambio, disponemos de un conjunto de etiquetas para cada instancia. Este problema está inspirado en el problema de la clasificación de neuronas, en el que un grupo de expertos proporciona una etiqueta de clase para cada instancia de manera individual. Se propone un método para aprender redes bayesianas utilizando vectores de cuentas, que representan el número de expertos que seleccionan cada etiqueta de clase para cada instancia. Estas redes bayesianas se evalúan utilizando bases de datos artificiales de problemas de aprendizaje supervisado.