Uncertainty Quantification with Support Vector Regression Prediction Models

Uncertainty quantification of petroleum reservoir models is one of the present challenges, which is usually approached with a wide range of geostatistical tools linked with statistical optimisation or/and inference algorithms. The paper considers a data driven approach in modelling uncertainty in spatial predictions. Proposed semi-supervised Support Vector Regression (SVR) model has demonstrated its capability to represent realistic features and describe stochastic variability and non-uniqueness of spatial properties. It is able to capture and preserve key spatial dependencies such as connectivity, which is often difficult to achieve with two-point geostatistical models. Semi-supervised SVR is designed to integrate various kinds of conditioning data and learn dependences from them. A stochastic semi-supervised SVR model is integrated into a Bayesian framework to quantify uncertainty with multiple models fitted to dynamic observations. The developed approach is illustrated with a reservoir case study. The resulting probabilistic production forecasts are described by uncertainty envelopes.

Transcriptional response to cardiac injury in the zebrafish: systematic identification of genes with highly concordant activity across in vivo models.

BACKGROUND: Zebrafish is a clinically-relevant model of heart regeneration. Unlike mammals, it has a remarkable heart repair capacity after injury, and promises novel translational applications. Amputation and cryoinjury models are key research tools for understanding injury response and regeneration in vivo. An understanding of the transcriptional responses following injury is needed to identify key players of heart tissue repair, as well as potential targets for boosting this property in humans. RESULTS: We investigated amputation and cryoinjury in vivo models of heart damage in the zebrafish through unbiased, integrative analyses of independent molecular datasets. To detect genes with potential biological roles, we derived computational prediction models with microarray data from heart amputation experiments. We focused on a top-ranked set of genes highly activated in the early post-injury stage, whose activity was further verified in independent microarray datasets. Next, we performed independent validations of expression responses with qPCR in a cryoinjury model. Across in vivo models, the top candidates showed highly concordant responses at 1 and 3 days post-injury, which highlights the predictive power of our analysis strategies and the possible biological relevance of these genes. Top candidates are significantly involved in cell fate specification and differentiation, and include heart failure markers such as periostin, as well as potential new targets for heart regeneration. For example, ptgis and ca2 were overexpressed, while usp2a, a regulator of the p53 pathway, was down-regulated in our in vivo models. Interestingly, a high activity of ptgis and ca2 has been previously observed in failing hearts from rats and humans. CONCLUSIONS: We identified genes with potential critical roles in the response to cardiac damage in the zebrafish. Their transcriptional activities are reproducible in different in vivo models of cardiac injury.

Genome-Wide Prediction Methods in Highly Diverse and Heterozygous Species: Proof-of-Concept through Simulation in Grapevine.

Nowadays, genome-wide association studies (GWAS) and genomic selection (GS) methods which use genome-wide marker data for phenotype prediction are of much potential interest in plant breeding. However, to our knowledge, no studies have been performed yet on the predictive ability of these methods for structured traits when using training populations with high levels of genetic diversity. Such an example of a highly heterozygous, perennial species is grapevine. The present study compares the accuracy of models based on GWAS or GS alone, or in combination, for predicting simple or complex traits, linked or not with population structure. In order to explore the relevance of these methods in this context, we performed simulations using approx 90,000 SNPs on a population of 3,000 individuals structured into three groups and corresponding to published diversity grapevine data. To estimate the parameters of the prediction models, we defined four training populations of 1,000 individuals, corresponding to these three groups and a core collection. Finally, to estimate the accuracy of the models, we also simulated four breeding populations of 200 individuals. Although prediction accuracy was low when breeding populations were too distant from the training populations, high accuracy levels were obtained using the sole core-collection as training population. The highest prediction accuracy was obtained (up to 0.9) using the combined GWAS-GS model. We thus recommend using the combined prediction model and a core-collection as training population for grapevine breeding or for other important economic crops with the same characteristics.

Quantifying rates of landscape evolution and tectonic processes by thermochronology and numerical modeling of crustal heat transport using PECUBE

PECUBE is a three-dimensional thermal-kinematic code capable of solving the heat production-diffusion-advection equation under a temporally varying surface boundary condition. It was initially developed to assess the effects of time-varying surface topography (relief) on low-temperature thermochronological datasets. Thermochronometric ages are predicted by tracking the time-temperature histories of rock-particles ending up at the surface and by combining these with various age-prediction models. In the decade since its inception, the PECUBE code has been under continuous development as its use became wider and addressed different tectonic-geomorphic problems. This paper describes several major recent improvements in the code, including its integration with an inverse-modeling package based on the Neighborhood Algorithm, the incorporation of fault-controlled kinematics, several different ways to address topographic and drainage change through time, the ability to predict subsurface (tunnel or borehole) data, prediction of detrital thermochronology data and a method to compare these with observations, and the coupling with landscape-evolution (or surface-process) models. Each new development is described together with one or several applications, so that the reader and potential user can clearly assess and make use of the capabilities of PECUBE. We end with describing some developments that are currently underway or should take place in the foreseeable future. (C) 2012 Elsevier B.V. All rights reserved.

Predicting smoking cessation and its relapse in HIV-infected patients: the Swiss HIV Cohort Study.

OBJECTIVES: The aim of the study was to assess whether prospective follow-up data within the Swiss HIV Cohort Study can be used to predict patients who stop smoking; or among smokers who stop, those who start smoking again. METHODS: We built prediction models first using clinical reasoning ('clinical models') and then by selecting from numerous candidate predictors using advanced statistical methods ('statistical models'). Our clinical models were based on literature that suggests that motivation drives smoking cessation, while dependence drives relapse in those attempting to stop. Our statistical models were based on automatic variable selection using additive logistic regression with component-wise gradient boosting. RESULTS: Of 4833 smokers, 26% stopped smoking, at least temporarily; because among those who stopped, 48% started smoking again. The predictive performance of our clinical and statistical models was modest. A basic clinical model for cessation, with patients classified into three motivational groups, was nearly as discriminatory as a constrained statistical model with just the most important predictors (the ratio of nonsmoking visits to total visits, alcohol or drug dependence, psychiatric comorbidities, recent hospitalization and age). A basic clinical model for relapse, based on the maximum number of cigarettes per day prior to stopping, was not as discriminatory as a constrained statistical model with just the ratio of nonsmoking visits to total visits. CONCLUSIONS: Predicting smoking cessation and relapse is difficult, so that simple models are nearly as discriminatory as complex ones. Patients with a history of attempting to stop and those known to have stopped recently are the best candidates for an intervention.

Antifungals in the ICU.

PURPOSE OF REVIEW: Invasive fungal infections remain a serious complication for critically ill ICU patients. The aim of this article is to review recent efficacy data of newer antifungal agents for the treatment of invasive candidiasis. The influence that recent epidemiological trends, advances in diagnostic testing, and risk prediction methods exert on the optimization of antifungal therapy for critically ill ICU patients will also be reviewed. RECENT FINDINGS: Recent clinical trials have documented the clinical efficacy of the echinocandins and the newer triazoles for the management of invasive candidiasis. Thus far, resistance to echinocandins remains rare. Changes in the epidemiology of Candida spp. causing invasive candidiasis, such as an increasing relative proportion of non-albicans Candida spp., have not been universally reported, although they have important implications for the use of fluconazole as first-line therapy for invasive candidiasis. Efforts to improve the timeliness and accuracy of laboratory diagnostic techniques and clinical prediction models to allow early and accurately targeted antifungal intervention strategies continue. SUMMARY: Echinocandins, given their clinical efficacy, spectrum of activity, and favourable pharmacological properties, are likely to replace fluconazole as initial antifungal agents of choice among critically ill ICU patients. The optimization of patient outcomes will require more accurately targeted early antifungal intervention strategies based upon sensitive and specific biological and clinical markers of risk.

Wavelet analysis residual kriging vs. neural network residual kriging

This paper deals with the problem of spatial data mapping. A new method based on wavelet interpolation and geostatistical prediction (kriging) is proposed. The method - wavelet analysis residual kriging (WARK) - is developed in order to assess the problems rising for highly variable data in presence of spatial trends. In these cases stationary prediction models have very limited application. Wavelet analysis is used to model large-scale structures and kriging of the remaining residuals focuses on small-scale peculiarities. WARK is able to model spatial pattern which features multiscale structure. In the present work WARK is applied to the rainfall data and the results of validation are compared with the ones obtained from neural network residual kriging (NNRK). NNRK is also a residual-based method, which uses artificial neural network to model large-scale non-linear trends. The comparison of the results demonstrates the high quality performance of WARK in predicting hot spots, reproducing global statistical characteristics of the distribution and spatial correlation structure.

Comparison of clinical and angiographic prognostic risk scores in elderly patients presenting with acute coronary syndrome and referred for percutaneous coronary intervention.

BACKGROUND: Multiple risk prediction models have been validated in all-age patients presenting with acute coronary syndrome (ACS) and treated with percutaneous coronary intervention (PCI); however, they have not been validated specifically in the elderly. METHODS: We calculated the GRACE (Global Registry of Acute Coronary Events) score, the logistic EuroSCORE, the AMIS (Acute Myocardial Infarction Swiss registry) score, and the SYNTAX (Synergy between Percutaneous Coronary Intervention with TAXUS and Cardiac Surgery) score in a consecutive series of 114 patients ≥75 years presenting with ACS and treated with PCI within 24 hours of hospital admission. Patients were stratified according to score tertiles and analysed retrospectively by comparing the lower/mid tertiles as an aggregate group with the higher tertile group. The primary endpoint was 30-day mortality. Secondary endpoints were the composite of death and major adverse cardiovascular events (MACE) at 30 days, and 1-year MACE-free survival. Model discrimination ability was assessed using the area under receiver operating characteristic curve (AUC). RESULTS: Thirty-day mortality was higher in the upper tertile compared with the aggregate lower/mid tertiles according to the logistic EuroSCORE (42% vs 5%; odds ratio [OR] = 14, 95% confidence interval [CI] = 4-48; p <0.001; AUC = 0.79), the GRACE score (40% vs 4%; OR = 17, 95% CI = 4-64; p <0.001; AUC = 0.80), the AMIS score (40% vs 4%; OR = 16, 95% CI = 4-63; p <0.001; AUC = 0.80), and the SYNTAX score (37% vs 5%; OR = 11, 95% CI = 3-37; p <0.001; AUC = 0.77). CONCLUSIONS: In elderly patients presenting with ACS and referred to PCI within 24 hours of admission, the GRACE score, the EuroSCORE, the AMIS score, and the SYNTAX score predicted 30 day mortality. The predictive value of clinical scores was improved by using them in combination.

Predictors of onset of cannabis and other drug use in male young adults: results from a longitudinal study.

BACKGROUND: The use of cannabis and other illegal drugs is particularly prevalent in male young adults and is associated with severe health problems. This longitudinal study explored variables associated with the onset of cannabis use and the onset of illegal drug use other than cannabis separately in male young adults, including demographics, religion and religiosity, health, social context, substance use, and personality. Furthermore, we explored how far the gateway hypothesis and the common liability to addiction model are in line with the resulting prediction models. METHODS: The data were gathered within the Cohort Study on Substance Use Risk Factors (C-SURF). Young men aged around 20 years provided demographic, social, health, substance use, and personality-related data at baseline. Onset of cannabis and other drug use were assessed at 15-months follow-up. Samples of 2,774 and 4,254 individuals who indicated at baseline that they have not used cannabis and other drugs, respectively, in their life and who provided follow-up data were used for the prediction models. Hierarchical logistic stepwise regressions were conducted, in order to identify predictors of the late onset of cannabis and other drug use separately. RESULTS: Not providing for oneself, having siblings, depressiveness, parental divorce, lower parental knowledge of peers and the whereabouts, peer pressure, very low nicotine dependence, and sensation seeking were positively associated with the onset of cannabis use. Practising religion was negatively associated with the onset of cannabis use. Onset of drug use other than cannabis showed a positive association with depressiveness, antisocial personality disorder, lower parental knowledge of peers and the whereabouts, psychiatric problems of peers, problematic cannabis use, and sensation seeking. CONCLUSIONS: Consideration of the predictor variables identified within this study may help to identify young male adults for whom preventive measures for cannabis or other drug use are most appropriate. The results provide evidence for both the gateway hypothesis and the common liability to addiction model and point to further variables like depressiveness or practising of religion that might influence the onset of drug use.

Kernel-based mapping of orographic rainfall enhancement in the Swiss Alps as detected by weatherradar

In this paper, we develop a data-driven methodology to characterize the likelihood of orographic precipitation enhancement using sequences of weather radar images and a digital elevation model (DEM). Geographical locations with topographic characteristics favorable to enforce repeatable and persistent orographic precipitation such as stationary cells, upslope rainfall enhancement, and repeated convective initiation are detected by analyzing the spatial distribution of a set of precipitation cells extracted from radar imagery. Topographic features such as terrain convexity and gradients computed from the DEM at multiple spatial scales as well as velocity fields estimated from sequences of weather radar images are used as explanatory factors to describe the occurrence of localized precipitation enhancement. The latter is represented as a binary process by defining a threshold on the number of cell occurrences at particular locations. Both two-class and one-class support vector machine classifiers are tested to separate the presumed orographic cells from the nonorographic ones in the space of contributing topographic and flow features. Site-based validation is carried out to estimate realistic generalization skills of the obtained spatial prediction models. Due to the high class separability, the decision function of the classifiers can be interpreted as a likelihood or susceptibility of orographic precipitation enhancement. The developed approach can serve as a basis for refining radar-based quantitative precipitation estimates and short-term forecasts or for generating stochastic precipitation ensembles conditioned on the local topography.

Kernel-based Mapping of Meteorological Fields in Complex Orography

Due to the advances in sensor networks and remote sensing technologies, the acquisition and storage rates of meteorological and climatological data increases every day and ask for novel and efficient processing algorithms. A fundamental problem of data analysis and modeling is the spatial prediction of meteorological variables in complex orography, which serves among others to extended climatological analyses, for the assimilation of data into numerical weather prediction models, for preparing inputs to hydrological models and for real time monitoring and short-term forecasting of weather.In this thesis, a new framework for spatial estimation is proposed by taking advantage of a class of algorithms emerging from the statistical learning theory. Nonparametric kernel-based methods for nonlinear data classification, regression and target detection, known as support vector machines (SVM), are adapted for mapping of meteorological variables in complex orography.With the advent of high resolution digital elevation models, the field of spatial prediction met new horizons. In fact, by exploiting image processing tools along with physical heuristics, an incredible number of terrain features which account for the topographic conditions at multiple spatial scales can be extracted. Such features are highly relevant for the mapping of meteorological variables because they control a considerable part of the spatial variability of meteorological fields in the complex Alpine orography. For instance, patterns of orographic rainfall, wind speed and cold air pools are known to be correlated with particular terrain forms, e.g. convex/concave surfaces and upwind sides of mountain slopes.Kernel-based methods are employed to learn the nonlinear statistical dependence which links the multidimensional space of geographical and topographic explanatory variables to the variable of interest, that is the wind speed as measured at the weather stations or the occurrence of orographic rainfall patterns as extracted from sequences of radar images. Compared to low dimensional models integrating only the geographical coordinates, the proposed framework opens a way to regionalize meteorological variables which are multidimensional in nature and rarely show spatial auto-correlation in the original space making the use of classical geostatistics tangled.The challenges which are explored during the thesis are manifolds. First, the complexity of models is optimized to impose appropriate smoothness properties and reduce the impact of noisy measurements. Secondly, a multiple kernel extension of SVM is considered to select the multiscale features which explain most of the spatial variability of wind speed. Then, SVM target detection methods are implemented to describe the orographic conditions which cause persistent and stationary rainfall patterns. Finally, the optimal splitting of the data is studied to estimate realistic performances and confidence intervals characterizing the uncertainty of predictions.The resulting maps of average wind speeds find applications within renewable resources assessment and opens a route to decrease the temporal scale of analysis to meet hydrological requirements. Furthermore, the maps depicting the susceptibility to orographic rainfall enhancement can be used to improve current radar-based quantitative precipitation estimation and forecasting systems and to generate stochastic ensembles of precipitation fields conditioned upon the orography.

Decision curve analysis revisited: overall net benefit, relationships to ROC curve analysis, and application to case-control studies.

ABSTRACT: BACKGROUND: Decision curve analysis has been introduced as a method to evaluate prediction models in terms of their clinical consequences if used for a binary classification of subjects into a group who should and into a group who should not be treated. The key concept for this type of evaluation is the "net benefit", a concept borrowed from utility theory. METHODS: We recall the foundations of decision curve analysis and discuss some new aspects. First, we stress the formal distinction between the net benefit for the treated and for the untreated and define the concept of the "overall net benefit". Next, we revisit the important distinction between the concept of accuracy, as typically assessed using the Youden index and a receiver operating characteristic (ROC) analysis, and the concept of utility of a prediction model, as assessed using decision curve analysis. Finally, we provide an explicit implementation of decision curve analysis to be applied in the context of case-control studies. RESULTS: We show that the overall net benefit, which combines the net benefit for the treated and the untreated, is a natural alternative to the benefit achieved by a model, being invariant with respect to the coding of the outcome, and conveying a more comprehensive picture of the situation. Further, within the framework of decision curve analysis, we illustrate the important difference between the accuracy and the utility of a model, demonstrating how poor an accurate model may be in terms of its net benefit. Eventually, we expose that the application of decision curve analysis to case-control studies, where an accurate estimate of the true prevalence of a disease cannot be obtained from the data, is achieved with a few modifications to the original calculation procedure. CONCLUSIONS: We present several interrelated extensions to decision curve analysis that will both facilitate its interpretation and broaden its potential area of application.

Premenopausal serum androgens and breast cancer risk: a nested case-control study.

ABSTRACT: INTRODUCTION: Prospective epidemiologic studies have consistently shown that levels of circulating androgens in postmenopausal women are positively associated with breast cancer risk. However, data in premenopausal women are limited. METHODS: A case-control study nested within the New York University Women's Health Study was conducted. A total of 356 cases (276 invasive and 80 in situ) and 683 individually-matched controls were included. Matching variables included age and date, phase, and day of menstrual cycle at blood donation. Testosterone, androstenedione, dehydroandrosterone sulfate (DHEAS) and sex hormone-binding globulin (SHBG) were measured using direct immunoassays. Free testosterone was calculated. RESULTS: Premenopausal serum testosterone and free testosterone concentrations were positively associated with breast cancer risk. In models adjusted for known risk factors of breast cancer, the odds ratios for increasing quintiles of testosterone were 1.0 (reference), 1.5 (95% confidence interval (CI), 0.9 to 2.3), 1.2 (95% CI, 0.7 to 1.9), 1.4 (95% CI, 0.9 to 2.3) and 1.8 (95% CI, 1.1 to 2.9; Ptrend = 0.04), and for free testosterone were 1.0 (reference), 1.2 (95% CI, 0.7 to 1.8), 1.5 (95% CI, 0.9 to 2.3), 1.5 (95% CI, 0.9 to 2.3), and 1.8 (95% CI, 1.1 to 2.8, Ptrend = 0.01). A marginally significant positive association was observed with androstenedione (P = 0.07), but no association with DHEAS or SHBG. Results were consistent in analyses stratified by tumor type (invasive, in situ), estrogen receptor status, age at blood donation, and menopausal status at diagnosis. Intra-class correlation coefficients for samples collected from 0.8 to 5.3 years apart (median 2 years) in 138 cases and 268 controls were greater than 0.7 for all biomarkers except for androstenedione (0.57 in controls). CONCLUSIONS: Premenopausal concentrations of testosterone and free testosterone are associated with breast cancer risk. Testosterone and free testosterone measurements are also highly reliable (that is, a single measurement is reflective of a woman's average level over time). Results from other prospective studies are consistent with our results. The impact of including testosterone or free testosterone in breast cancer risk prediction models for women between the ages of 40 and 50 years should be assessed. Improving risk prediction models for this age group could help decision making regarding both screening and chemoprevention of breast cancer.

The ADO Index as a Predictor of Two-Year Mortality in General Practice-Based Chronic Obstructive Pulmonary Disease Cohorts.

BACKGROUND: Existing prediction models for mortality in chronic obstructive pulmonary disease (COPD) patients have not yet been validated in primary care, which is where the majority of patients receive care. OBJECTIVES: Our aim was to validate the ADO (age, dyspnoea, airflow obstruction) index as a predictor of 2-year mortality in 2 general practice-based COPD cohorts. METHODS: Six hundred and forty-six patients with COPD with GOLD (Global Initiative for Chronic Obstructive Lung Disease) stages I-IV were enrolled by their general practitioners and followed for 2 years. The ADO regression equation was used to predict a 2-year risk of all-cause mortality in each patient and this risk was compared with the observed 2-year mortality. Discrimination and calibration were assessed as well as the strength of association between the 15-point ADO score and the observed 2-year all-cause mortality. RESULTS: Fifty-two (8.1%) patients died during the 2-year follow-up period. Discrimination with the ADO index was excellent with an area under the curve of 0.78 [95% confidence interval (CI) 0.71-0.84]. Overall, the predicted and observed risks matched well and visual inspection revealed no important differences between them across 10 risk classes (p = 0.68). The odds ratio for death per point increase according to the ADO index was 1.50 (95% CI 1.31-1.71). CONCLUSIONS: The ADO index showed excellent prediction properties in an out-of-population validation carried out in COPD patients from primary care settings. © 2014 S. Karger AG, Basel.

Comparison of 2 frailty indexes for prediction of falls, disability, fractures, and death in older women.

BACKGROUND: Frailty, as defined by the index derived from the Cardiovascular Health Study (CHS index), predicts risk of adverse outcomes in older adults. Use of this index, however, is impractical in clinical practice. METHODS: We conducted a prospective cohort study in 6701 women 69 years or older to compare the predictive validity of a simple frailty index with the components of weight loss, inability to rise from a chair 5 times without using arms, and reduced energy level (Study of Osteoporotic Fractures [SOF index]) with that of the CHS index with the components of unintentional weight loss, poor grip strength, reduced energy level, slow walking speed, and low level of physical activity. Women were classified as robust, of intermediate status, or frail using each index. Falls were reported every 4 months for 1 year. Disability (> or =1 new impairment in performing instrumental activities of daily living) was ascertained at 4(1/2) years, and fractures and deaths were ascertained during 9 years of follow-up. Area under the curve (AUC) statistics from receiver operating characteristic curve analysis and -2 log likelihood statistics were compared for models containing the CHS index vs the SOF index. RESULTS: Increasing evidence of frailty as defined by either the CHS index or the SOF index was similarly associated with an increased risk of adverse outcomes. Frail women had a higher age-adjusted risk of recurrent falls (odds ratio, 2.4), disability (odds ratio, 2.2-2.8), nonspine fracture (hazard ratio, 1.4-1.5), hip fracture (hazard ratio, 1.7-1.8), and death (hazard ratio, 2.4-2.7) (P < .001 for all models). The AUC comparisons revealed no differences between models with the CHS index vs the SOF index in discriminating falls (AUC = 0.61 for both models; P = .66), disability (AUC = 0.64; P = .23), nonspine fracture (AUC = 0.55; P = .80), hip fracture (AUC = 0.63; P = .64), or death (AUC = 0.72; P = .10). Results were similar when -2 log likelihood statistics were compared. CONCLUSION: The simple SOF index predicts risk of falls, disability, fracture, and death as well as the more complex CHS index and may provide a useful definition of frailty to identify older women at risk of adverse health outcomes in clinical practice.