Janos Hajagos

Linked open drug data for pharmaceutical research and development

There is an abundance of information about drugs available on the Web. Data sources range from medicinal chemistry results, over the impact of drugs on gene expression, to the outcomes of drugs in clinical trials. These data are typically not connected together, which reduces the ease with which insights can be gained. Linking Open Drug Data (LODD) is a task force within the World Wide Web Consortiums (W3C) Health Care and Life Sciences Interest Group (HCLS IG). LODD has surveyed publicly available data about drugs, created Linked Data representations of the data sets, and identified interesting scientific and business questions that can be answered once the data sets are connected. The task force provides recommendations for the best practices of exposing data in a Linked Data representation. In this paper, we present past and ongoing work of LODD and discuss the growing importance of Linked Data as a foundation for pharmaceutical R&D data sharing.

Experimental Uncertainty Estimation and Statistics for Data Having Interval Uncertainty

This report addresses the characterization of measurements that include epistemic uncertainties in the form of intervals. It reviews the application of basic descriptive statistics to data sets which contain intervals rather than exclusively point estimates. It describes algorithms to compute various means, the median and other percentiles, variance, interquartile range, moments, confidence limits, and other important statistics and summarizes the computability of these statistics as a function of sample size and characteristics of the intervals in the data (degree of overlap, size and regularity of widths, etc.). It also reviews the prospects for analyzing such data sets with the methods of inferential statistics such as outlier detection and regressions. The report explores the tradeoff between measurement precision and sample size in statistical results that are sensitive to both. It also argues that an approach based on interval statistics could be a reasonable alternative to current standard methods for evaluating, expressing and propagating measurement uncertainties.

Arithmetic with uncertain numbers: rigorous and (often) best possible answers

Abstract A variety of complex arithmetic problems can be solved using a single—and fairly simple—approach based on probability bounds analysis. The inputs are first expressed as interval bounds on cumulative distribution functions. Each uncertain input variable is then decomposed into a list of pairs of the form (interval, probability). A Cartesian product of these lists, reflecting both the independence among inputs and the mathematical expression that binds them together, creates another list, which is recomposed to form the resulting uncertain number as upper and lower bounds on a cumulative distribution function. Ancillary techniques are also employed, such as condensation, which is necessary to keep the length of the list from growing inordinately in sequential operations, and subinterval reconstitution, which is needed to solve interval arithmetic problems involving repeated parameters. Moment propagation formulas are simultaneously used to bound mean and variance estimates accompanying the bounds on the cumulative distribution function. Generalizations of this approach are also described that allow for dependencies other than independence, completely unknown dependence, and model uncertainty more generally.

Dependence in probabilistic modeling, Dempster-Shafer theory, and probability bounds analysis.

This report summarizes methods to incorporate information (or lack of information) about inter-variable dependence into risk assessments that use Dempster-Shafer theory or probability bounds analysis to address epistemic and aleatory uncertainty. The report reviews techniques for simulating correlated variates for a given correlation measure and dependence model, computation of bounds on distribution functions under a specified dependence model, formulation of parametric and empirical dependence models, and bounding approaches that can be used when information about the intervariable dependence is incomplete. The report also reviews several of the most pervasive and dangerous myths among risk analysts about dependence in probabilistic models.

Towards Combining Probabilistic and Interval Uncertainty in Engineering Calculations: Algorithms for Computing Statistics under Interval Uncertainty, and Their Computational Complexity

In many engineering applications, we have to combine probabilistic and interval uncertainty. For example, in environmental analysis, we observe a pollution level x(t) in a lake at different moments of time t, and we would like to estimate standard statistical characteristics such as mean, variance, autocorrelation, correlation with other measurements. In environmental measurements, we often only measure the values with interval uncertainty. We must therefore modify the existing statistical algorithms to process such interval data.In this paper, we provide a survey of algorithms for computing various statistics under interval uncertainty and their computational complexity. The survey includes both known and new algorithms.

Varying correlation coefficients can underestimate uncertainty in probabilistic models

Abstract In accounting for the dependencies among variables in probabilistic (convolution) models, a sensitivity study that varies a correlation between plausible values, even the extremes of +1 and −1, cannot characterize the possible range of results that could be entailed by nonlinear dependencies. Because a functional modeling strategy that seeks to model mechanistically the underlying sources of the dependencies will often be untenable, a phenomenological approach will often be needed to handle dependencies. We summarize recent algorithmic advances that allow the calculation of results under particular bivariate dependence functions, under only partially specified dependence functions, or even without any assumption whatever about dependence.

Constructor:synthesizing information about uncertain variables.

Constructor is software for the Microsoft Windows microcomputer environment that facilitates the collation of empirical information and expert judgment for the specification of probability distributions, probability boxes, random sets or Dempster-Shafer structures from data, qualitative shape information, constraints on moments, order statistics, densities, and coverage probabilities about uncertain unidimensional quantities. These quantities may be real-valued, integer-valued or logical values.

Interval Monte Carlo as an Alternative to Second-Order Sampling for Estimating Ecological Risk

Interval Monte Carlo offers an alternative to second-order approaches for modeling measurement uncertainty in a simulation framework. Using the example of computing quasi-extinction decline risk for an ecological population, an interval Monte Carlo model is built. If the model is not written optimally, the mean and standard deviation of the growth rate repeat, then the bounds on the quasi-extinction risk will be sub-optimal. Depending on your operational definition of what an interval is, the sub-optimal bounds may be the best possible bounds. A comparison between second-order and interval Monte Carlo is made, which reveals that second-order approaches can underestimate the upper bound on the quasi-extinction decline risk to the population when there are a large number of parameters that need to be sampled.

Evaluation of VTE prophylaxis and the impact of alternate regimens on post-operative bleeding and thrombotic complications following bariatric procedures

BackgroundStudies examining utilization and impact of venous thromboembolism (VTE) chemoprophylaxis for patients undergoing bariatric surgery are limited. Determination of the optimal prophylactic regimen to minimize complications is crucial.MethodsThe Cerner Health Facts database from 2003 to 2013 was queried using ICD-9 codes to identify patients undergoing laparoscopic sleeve gastrectomy (LSG) and Roux-en-Y gastric bypass (RYGB). VTE chemoprophylaxis regimens were divided into pre-operative alone (PreP), post-operative alone (PostP), both pre-operative and post-operative (PPP), or no prophylaxis (NP). Specific chemoprophylaxis agents were compared. Comparisons in inpatient clinical outcomes were based on univariate analysis and multivariable logistic regression when appropriate.ResultsWe identified 11,860 patients who underwent LSG and RYGB. 634 (5.35%) had PreP, 4593 (38.73%) had PostP, 2646 (22.31%) had PPP, and 3987 (33.62%) had NP. The overall rates of transfusion, DVT, and PE were 2.48, 0.27, and 0.18%, respectively. Patients without chemoprophylaxis had higher rate of DVT compared to any chemoprophylaxis (0.58 vs 0.11%, p < 0.0001), without any significant difference in PE rate. Patients with pre-operative chemoprophylaxis were more likely to receive transfusion compared to patients with post-operative prophylaxis alone (OR 1.98, 95% CI 1.28–3), without significant difference in having VTE. When examining heparin versus enoxaparin versus mixed regimen in the PostP group, mixed regimen was associated with increased transfusion requirements (p < 0.001).ConclusionsBariatric surgical VTE chemoprophylaxis utilization is inconsistent. In this study, post-operative VTE chemoprophylaxis was associated with decreased VTE events compared to NP, while minimizing bleeding compared to PreP. Mixed therapy using heparin and enoxaparin was associated with more bleeding.

Linking Clinicians to Biomedical Researchers: An Application of the ISF Ontology at Stony Brook Medicine

From an administrative system, clinical experience was computed for 30 clinical faculty at Stony Brook Medicine. Experience was computed and represented in the ISF ontology using ICD9CM diagnosis codes that were mapped to MeSH. Using SPARQL, a semantic web technology for data querying, and data from a VIVO site, it was possible to connect to 257 different biomedical researchers based on published works.