Developing computational model-based diagnostics to analyse clinical chemistry data.

This article provides methodological and technical considerations to researchers starting to develop computational model-based diagnostics using clinical chemistry data. These models are of increasing importance, since novel metabolomics and proteomics measuring technologies are able to produce large amounts of data that are difficult to interpret at first sight, but have high diagnostic potential. Computational models aid interpretation and make the data accessible for clinical diagnosis. We discuss the issues that a modeller has to take into account during the design, construction and evaluation phases of model development. We use the example of Particle Profiler development, a model-based diagnostic tool for lipoprotein disorders, as a case study, to illustrate our considerations. The case study also offers techniques for efficient model formulation, model calculation, workflow structuring and quality control.

Systematic construction of a conceptual minimal model of plasma cholesterol levels based on knockout mouse phenotypes.

Elevated plasma cholesterol, a well-known risk factor for cardiovascular diseases, is the result of the activity of many genes and their encoded proteins in a complex physiological network. We aim to develop a minimal kinetic computational model for predicting plasma cholesterol levels. To define the scope of this model, it is essential to discriminate between important and less important processes influencing plasma cholesterol levels. To this end, we performed a systematic review of mouse knockout strains and used the resulting dataset, named KOMDIP, for the identification of key genes that determine plasma cholesterol levels. Based on the described phenotype of mouse knockout models, 36 of the 120 evaluated genes were marked as key genes that have a pronounced effect on the plasma cholesterol concentration. The key genes include well-known genes, e.g., Apoe and Ldlr, as well as genes hardly linked to cholesterol metabolism so far, e.g., Plagl2 and Slc37a4. Based on the catalytic function of the genes, a minimal conceptual model was defined. A comparison with nine conceptual models from literature revealed that each of the individual published models is less complete than our model. Concluding, we have developed a conceptual model that can be used to develop a physiologically based kinetic model to quantitatively predict plasma cholesterol levels.

Nutritional systems biology modeling: from molecular mechanisms to physiology.

The use of computational modeling and simulation has increased in many biological fields, but despite their potential these techniques are only marginally applied in nutritional sciences. Nevertheless, recent applications of modeling have been instrumental in answering important nutritional questions from the cellular up to the physiological levels. Capturing the complexity of today's important nutritional research questions poses a challenge for modeling to become truly integrative in the consideration and interpretation of experimental data at widely differing scales of space and time. In this review, we discuss a selection of available modeling approaches and applications relevant for nutrition. We then put these models into perspective by categorizing them according to their space and time domain. Through this categorization process, we identified a dearth of models that consider processes occurring between the microscopic and macroscopic scale. We propose a "middle-out" strategy to develop the required full-scale, multilevel computational models. Exhaustive and accurate phenotyping, the use of the virtual patient concept, and the development of biomarkers from "-omics" signatures are identified as key elements of a successful systems biology modeling approach in nutrition research--one that integrates physiological mechanisms and data at multiple space and time scales.

New in vitro dermal absorption database and the prediction of dermal absorption under finite conditions for risk assessment purposes.

Most QSARs for dermal absorption predict the permeability coefficient, K(p), of a molecule, which is valid for infinite dose conditions. In practice, dermal exposure mostly occurs under finite dose conditions. Therefore, a simple model to predict finite dose dermal absorption from infinite dose data (K(p) and lag time) and the stratum corneum/water partition coefficient (K(SC,W)) was developed. To test the model, a series of in vitro dermal absorption experiments was performed under both infinite and finite dose conditions using acetic acid, benzoic acid, bis(2-ethylhexyl)phthalate, butoxyethanol, cortisone, decanol, diazinone, 2,4-dichlorophenol, ethacrynic acid, linolenic acid, octylparaben, oleic acid, propylparaben, salicylic acid and testosterone. For six substances, the predicted relative dermal absorption was not statistically different from the measured value. For all other substances, measured absorption was overpredicted by the model, but most of the overpredictions were still below the European default absorption value. In conclusion, our finite dose prediction model provides a useful and cost-effective estimate of dermal absorption, to be used in risk assessment for non-volatile substances dissolved in water at non-irritating concentrations.

Improved cholesterol phenotype analysis by a model relating lipoprotein life cycle processes to particle size.

Increased plasma cholesterol is a known risk factor for cardiovascular disease. Lipoprotein particles transport both cholesterol and triglycerides through the blood. It is thought that the size distribution of these particles codetermines cardiovascular disease risk. New types of measurements can determine the concentration of many lipoprotein size-classes but exactly how each small class relates to disease risk is difficult to clear up. Because relating physiological process status to disease risk seems promising, we propose investigating how lipoprotein production, lipolysis, and uptake processes depend on particle size. To do this, we introduced a novel model framework (Particle Profiler) and evaluated its feasibility. The framework was tested using existing stable isotope flux data. The model framework implementation we present here reproduced the flux data and derived lipoprotein size pattern changes that corresponded to measured changes. It also sensitively indicated changes in lipoprotein metabolism between patient groups that are biologically plausible. Finally, the model was able to reproduce the cholesterol and triglyceride phenotype of known genetic diseases like familial hypercholesterolemia and familial hyperchylomicronemia. In the future, Particle Profiler can be applied for analyzing detailed lipoprotein size profile data and deriving rates of various lipolysis and uptake processes if an independent production estimate is given.

A physiologically based biokinetic (PBBK) model for estragole bioactivation and detoxification in rat.

The present study defines a physiologically based biokinetic (PBBK) model for the alkenylbenzene estragole in rat based on in vitro metabolic parameters determined using relevant tissue fractions, in silico derived partition coefficients, and physiological parameters derived from the literature. The model consists of eight compartments including liver, lung and kidney as metabolizing compartments, and additional compartments for fat, arterial blood, venous blood, rapidly perfused tissue and slowly perfused tissue. Evaluation of the model was performed by comparing the PBBK predicted dose-dependent formation of the estragole metabolites 4-allylphenol and 1'-hydroxyestragole glucuronide to literature reported levels of these metabolites, which were demonstrated to be in the same order of magnitude. With the model obtained the relative extent of bioactivation and detoxification of estragole at different oral doses was examined. At low doses formation of 4-allylphenol, leading to detoxification, is observed to be the major metabolic pathway, occurring mainly in the lung and kidney due to formation of this metabolite with high affinity in these organs. Saturation of this metabolic pathway in the lung and kidney leads to a relative increase in formation of the proximate carcinogenic metabolite 1'-hydroxyestragole, occurring mainly in the liver. This relative increase in formation of 1'-hydroxyestragole leads to a relative increase in formation of 1'-hydroxyestragole glucuronide and 1'-sulfooxyestragole the latter being the ultimate carcinogenic metabolite of estragole. These results indicate that the relative importance of different metabolic pathways of estragole may vary in a dose-dependent way, leading to a relative increase in bioactiviation of estragole at higher doses.

A network biology model of micronutrient related health.

Micronutrients are involved in specific biochemical pathways and have dedicated functions in the body, but they are also interconnected in complex metabolic networks, such as oxidative-reductive and inflammatory pathways and hormonal regulation, in which the overarching function is to optimise health. Post-genomic technologies, in particular metabolomics and proteomics, both of which are appropriate for plasma samples, provide a new opportunity to study the metabolic effects of micronutrients in relation to optimal health. The study of micronutrient-related health status requires a combination of data on markers of dietary exposure, markers of target function and biological response, health status metabolites, and disease parameters. When these nutrient-centred and physiology/health-centred parameters are combined and studied using a systems biology approach with bioinformatics and multivariate statistical tools, it should be possible to generate a micronutrient phenotype database. From this we can explore external factors that define the phenotype, such as lifestage and lifestyle, and the impact of genotype, and the results can also be used to define micronutrient requirements and provide dietary advice. New mechanistic insights have already been developed using biological network models, for example genes and protein-protein interactions in the aetiology of type 2 diabetes mellitus. It is hoped that the challenge of applying this approach to micronutrients will, in time, result in a change from micronutrient oriented to a health oriented views and provide a more holistic understanding of the role played by multiple micronutrients in the maintenance of homeostasis and prevention of chronic disease, for example through their involvement in oxidation and inflammation.

Quantitative profiling of bile acids in biofluids and tissues based on accurate mass high resolution LC-FT-MS: compound class targeting in a metabolomics workflow.

We report a sensitive, generic method for quantitative profiling of bile acids and other endogenous metabolites in small quantities of various biological fluids and tissues. The method is based on a straightforward sample preparation, separation by reversed-phase high performance liquid-chromatography mass spectrometry (HPLC-MS) and electrospray ionisation in the negative ionisation mode (ESI-). Detection is performed in full scan using the linear ion trap Fourier transform mass spectrometer (LTQ-FTMS) generating data for many (endogenous) metabolites, not only bile acids. A validation of the method in urine, plasma and liver was performed for 17 bile acids including their taurine, sulfate and glycine conjugates. The method is linear in the 0.01-1 microM range. The accuracy in human plasma ranges from 74 to 113%, in human urine 77 to 104% and in mouse liver 79 to 140%. The precision ranges from 2 to 20% for pooled samples even in studies with large number of samples (n>250). The method was successfully applied to a multi-compartmental APOE*3-Leiden mouse study, the main goal of which was to analyze the effect of increasing dietary cholesterol concentrations on hepatic cholesterol homeostasis and bile acid synthesis. Serum and liver samples from different treatment groups were profiled with the new method. Statistically significant differences between the diet groups were observed regarding total as well as individual bile acid concentrations.

Quercetin increases the bioavailability of 2-amino-1-methyl-6-phenylimidazo[4,5-b]pyridine (PhIP) in rats.

This study investigates whether the previous observation that quercetin increases the transport of PhIP through Caco-2 monolayers in vitro could be confirmed in an in vivo rat model. Co-administration of 1.45 micromol PhIP/kg bw and 30 micromol quercetin/kg bw significantly increased the blood AUC(0-8h) of PhIP in rats to 131+/-14% of the AUC(0-8h) for rats dosed with PhIP alone. Significantly increased blood PhIP levels were detected at 15, 30, 45 and 180 min. At 4 and 8h post-dosing a difference in the PhIP levels in the blood between the two treatment groups was no longer observed. In vitro and in silico modeling of PhIP transport using Caco-2 cells and a previously described kinetic model for PhIP transport revealed that the relative increase in PhIP transport caused by quercetin is dependent on the concentration of the two compounds. When substituting the PhIP and quercetin concentrations used in the in vivo experiment in the kinetic model, an effect of quercetin on PhIP transport was predicted that matches the actual effect of 131% observed in vivo. It is concluded that quercetin increases the bioavailability of the pro-carcinogen PhIP in rats pointing at a potential adverse effect of this supposed beneficial food ingredient.

Prediction of in vivo embryotoxic effect levels with a combination of in vitro studies and PBPK modelling.

The new EU legislations for chemicals (Registration, Evaluation and Authorization of Chemicals, REACH) and cosmetics (Seventh Amendment) stimulate the acceptance of in vitro and in silico approaches to test chemicals for their potential to cause reproductive effects. In the current study seven compounds with known in vivo developmental effects were tested in the embryonic stem cell test (EST). The EST correctly classified 5-fluorouracil, methotrexate, retinoic acid, 2-ethoxyacetic acid and 2-methoxyacetic acid for their in vivo embryotoxic potential. The toxicity of 2-methoxyethanol and 2-ethoxyethanol was underestimated due to a lack of metabolic capacity in the EST. This study further investigated the possibility to use in silico techniques to extrapolate in vitro effect concentrations determined in the EST to in vivo exposure levels. This approach was evaluated by comparing in silico predicted in vivo effect levels with effect levels measured in rodents. The in vivo effect levels of 2-methoxyethanol, 2-ethoxyethanol, methotrexate and retinoic acid were correctly predicted with in silico modelling. Contrary, in vivo embryotoxicity of 5-fluorouracil was overestimated following this approach. It is concluded that a combination of in vitro and in silico techniques appears to be a promising alternative test method for risk assessment of embryotoxic compounds.

Association between consumption of cruciferous vegetables and condiments and excretion in urine of isothiocyanate mercapturic acids.

A high intake of cruciferous vegetables is associated with a reduced risk of cancer and cardiovascular diseases. This protective effect has been linked to isothiocyanates, enzymatic hydrolysis products of glucosinolates. In this study, the metabolic fate of glucosinolates and isothiocyanates after ingestion of 19 different cruciferous vegetables was studied in three male subjects. After the consumption of 13 cruciferous vegetables (glucosinolate content, 0.01-0.94 mmol/kg) and six condiments (isothiocyanate content, 0.06-49.3 mmol/kg), eight different isothiocyanate mercapturic acids were determined in urine samples. Excretion levels after the consumption of raw vegetables and condiments were higher (bioavailability, 8.2-113%) as compared to cooked vegetables (bioavailability, 1.8-43%), but the excretion rate was similar (t1/2=2.1-3.9 h). Isothiocyanates in urine remain longer at a nonzero level after the consumption of glucosinolates from cooked vegetables, as compared to raw vegetables and condiments, and maximal levels in urine were reached about 4 h later. Isothiocyanate mercapturic acids can be used as a biomarker to reflect the active dose of isothiocyanates absorbed.

Quantum chemistry based quantitative structure-activity relationships for modeling the (sub)acute toxicity of substituted mononitrobenzenes in aquatic systems.

Fifteen experimental literature data sets on the acute toxicity of substituted nitrobenzenes to algae (Scenedesmus obliquus, Chlorella pyrenoidosa, C. vulgaris), daphnids (Daphnia magna, D. carinata), fish (Cyprinus carpio, Poecilia reticulata), protozoa (Tetrahymena pyriformis), bacteria (Phosphobacterium phosphoreum), and yeast (Saccharomyces cerevisiae) were used to establish quantum chemistry based quantitative structure-activity relationships (QSARs). The logarithm of the octanol/water partition coefficient, log Kow, and the energy of the lowest unoccupied molecular orbital, Elumo, were used as descriptors. Suitable QSAR models (0.65 < r2 < 0.98) to predict acute toxicity of substituted mononitrobenzenes to protozoa, fish, daphnids, yeast, and algae have been derived. The log Kow was a sufficient descriptor for all cases, with the additional Elumo descriptor being required only for algae. The QSARs were found to be valid for neutral substituted mononitrobenzenes with no -OH, -COOH, or -CN substituents attached directly to the ring. From the 100,196 European Inventory of Existing Commercial Substances (EINECS), 497 chemicals were identified that fit the selection criteria for the established QSARs. Based on these results, an advisory tool has been developed that directs users to the appropriate QSAR model to apply for various types of organisms within specified log Kow ranges. Using this tool, it is possible to obtain a good indication of the toxicity of a large set of EINECS chemicals and newly developed substituted mononitrobenzenes to five different organisms without the need for additional experimental testing.

Neurobehavioural evaluation and kinetics of inhalation of constant or fluctuating toluene concentrations in human volunteers.

The health risks of inhalation exposure to volatile organic solvents may not only depend on the total external dose, but also on the pattern of exposure. It has been suggested that exposure to regularly occurring peak concentrations may have a stronger impact on the brain than constant exposure at the same average level. Recent animal experimental studies conducted in our laboratory using relatively high concentrations of toluene have shown different effects on discrimination performance and motor activity during and after exposure, depending on the exposure scenario. Relevance of these findings for man was evaluated in a volunteer study in which 11 healthy men (age 20-49 years) were exposed by inhalation for 4h to either a constant concentration of 40ppm toluene or to three 30-min exposure peaks at 110ppm during this 4h period. Selected tests from the Neurobehavioural Evaluation System (NES) were performed repeatedly during and after exposure. Blood concentrations of toluene as well as urinary o-cresol excretion were measured at relevant time points. The results show that toluene concentration in blood increased during constant exposure and fluctuated during occupationally relevant peak exposures. Presumably, brain concentrations showed similar qualitative patterns. No clear changes were observed on neurobehavioural measures of motor performance, attention, perceptual coding and memory, or on measures of mood and affect. The exposure conditions do not seem to induce significant acute changes in central nervous system function similar to those observed at much higher concentrations in animals, although a statistical correlation was found between one motor performance test (Finger Tapping Test with alternating hands) and blood toluene concentrations. Urinary o-cresol excretion appeared to be significantly higher during the first 2h after exposure.

Toxicology of chemical mixtures: a challenging quest along empirical sciences.

This paper describes the "quest" of our institute trying to assess the toxicology of chemical mixtures. In this overview, we will discuss some critical developments in hazard identification and risk assessment of chemical mixtures during these past 15 years. We will stand still at empirical and mechanistic modeling. "Empirical" means that only information on doses or concentrations and effects is available in addition to an often empirically selected quantitative dose-response relationship. Empirical models have played a dominant role in the last decade to identify health and safety characteristics of chemical mixtures. Many of these models are based on the work of pioneers in mixture toxicology who defined three basic types of action for combinations of chemicals: simple similar action, simple dissimilar action and interaction. Nowadays, empirical models are mainly based on response-surface analysis and make use of advanced statistical designs. However, possible interactions between components in a mixture can also be given in terms of mechanistic models. In terms of "mechanistic" (or biological) understanding, interactions between compounds may occur in the kinetic phase (processes of uptake, distribution, metabolism and excretion) or in the dynamic phase (effects of chemicals on the receptor, cellular target or organ). A biological phenomenon such as competitive agonism as described for mixtures of drugs (biotransformation enzymes) or sensory irritants (nerve receptors) can accurately predict the effect of any of these mixtures. Thus, far mechanistic and empirical analyses of interactions are usually unrelated. It is one of the future challenges for mixtures research to combine information from both approaches. Also, our current biology-based models have their limitations, since they cannot integrate every relevant biological mechanism. In this respect, mechanistic modeling of mixtures may benefit from the developments coming from the arena of molecular biology (toxicogenomics) which offers an in-depth analysis of several involved enzymatic pathways in parallel through the use of a systems biology approach. This was illustrated with mixtures of food additives known to affect the liver. Key to further maturation of mixture toxicology is collaboration of experimental toxicologists, biomathematicians, biologists, pharmacologists, model developers, molecular biologists and bioinformaticians to ensure parallel and coordinated research in this challenging area of toxicology. For this reason, the next sequel will be even more challenging and exciting to that first 15 years of empirical testing.

The use of in vitro metabolic parameters and physiologically based pharmacokinetic (PBPK) modeling to explore the risk assessment of trichloroethylene.

A physiologically based pharmacokinetic (PBPK) model has been developed for trichloroethylene (1,1,2-trichloroethene, TRI) for rat and humans, based on in vitro metabolic parameters. These were obtained using individual cytochrome P450 and glutathione S-transferase enzymes. The main enzymes involved both for rats and humans are CYP2E1 and the µ- and π-class glutathione S-transferases. Validation experiments were performed in order to test the predictive value of the enzyme kinetic parameters to describe 'whole-body' disposition. Male Wistar rats were dosed orally or intravenously with different doses of trichloroethylene. Obtained exhaled radioactivity, excreted radioactivity in urine, and obtained blood concentration-time curves of trichloroethylene for all dosing groups were compared to predictions from the PBPK model. Subsequently, using the scaling factor derived from the rat experiments predictions were made for the extreme cases to be expected in humans, based on interindividual variations of the key enzymes involved. On comparing these predictions with literature data a very close match was found. This illustrates the potential application of in vitro metabolic parameters in risk assessment, through the use of PBPK modeling as a tool to understand and predict in vivo data. From a hypothetical 8 h exposure scenario to 35 ppm trichloroethylene in rats and humans, and assuming that the glutathione S-transferase pathway is responsible for the toxicity of trichloroethylene, it was concluded that humans are less sensitive for trichloroethylene toxicity than rats.

Dietary medium chain fatty acid supplementation leads to reduced VLDL lipolysis and uptake rates in comparison to linoleic acid supplementation.

Dietary medium chain fatty acids (MCFA) and linoleic acid follow different metabolic routes, and linoleic acid activates PPAR receptors. Both these mechanisms may modify lipoprotein and fatty acid metabolism after dietary intervention. Our objective was to investigate how dietary MCFA and linoleic acid supplementation and body fat distribution affect the fasting lipoprotein subclass profile, lipoprotein kinetics, and postprandial fatty acid kinetics. In a randomized double blind cross-over trial, 12 male subjects (age 51±7 years; BMI 28.5±0.8 kg/m2), were divided into 2 groups according to waist-hip ratio. They were supplemented with 60 grams/day MCFA (mainly C8:0, C10:0) or linoleic acid for three weeks, with a wash-out period of six weeks in between. Lipoprotein subclasses were measured using HPLC. Lipoprotein and fatty acid metabolism were studied using a combination of several stable isotope tracers. Lipoprotein and tracer data were analyzed using computational modeling. Lipoprotein subclass concentrations in the VLDL and LDL range were significantly higher after MCFA than after linoleic acid intervention. In addition, LDL subclass concentrations were higher in lower body obese individuals. Differences in VLDL metabolism were found to occur in lipoprotein lipolysis and uptake, not production; MCFAs were elongated intensively, in contrast to linoleic acid. Dietary MCFA supplementation led to a less favorable lipoprotein profile than linoleic acid supplementation. These differences were not due to elevated VLDL production, but rather to lower lipolysis and uptake rates.

Diagnostic markers based on a computational model of lipoprotein metabolism.

BACKGROUND: Dyslipidemia is an important risk factor for cardiovascular disease and type II diabetes. Lipoprotein diagnostics, such as LDL cholesterol and HDL cholesterol, help to diagnose these diseases. Lipoprotein profile measurements could improve lipoprotein diagnostics, but interpretational complexity has limited their clinical application to date. We have previously developed a computational model called Particle Profiler to interpret lipoprotein profiles. In the current study we further developed and calibrated Particle Profiler using subjects with specific genetic conditions. We subsequently performed technical validation and worked at an initial indication of clinical usefulness starting from available data on lipoprotein concentrations and metabolic fluxes. Since the model outcomes cannot be measured directly, the only available technical validation was corroboration. For an initial indication of clinical usefulness, pooled lipoprotein metabolic flux data was available from subjects with various types of dyslipidemia. Therefore we investigated how well lipoprotein metabolic ratios derived from Particle Profiler distinguished reported dyslipidemic from normolipidemic subjects. RESULTS: We found that the model could fit a range of normolipidemic and dyslipidemic subjects from fifteen out of sixteen studies equally well, with an average 8.8% ± 5.0% fit error; only one study showed a larger fit error. As initial indication of clinical usefulness, we showed that one diagnostic marker based on VLDL metabolic ratios better distinguished dyslipidemic from normolipidemic subjects than triglycerides, HDL cholesterol, or LDL cholesterol. The VLDL metabolic ratios outperformed each of the classical diagnostics separately; they also added power of distinction when included in a multivariate logistic regression model on top of the classical diagnostics. CONCLUSIONS: In this study we further developed, calibrated, and corroborated the Particle Profiler computational model using pooled lipoprotein metabolic flux data. From pooled lipoprotein metabolic flux data on dyslipidemic patients, we derived VLDL metabolic ratios that better distinguished normolipidemic from dyslipidemic subjects than standard diagnostics, including HDL cholesterol, triglycerides and LDL cholesterol. Since dyslipidemias are closely linked to cardiovascular disease and diabetes type II development, lipoprotein metabolic ratios are candidate risk markers for these diseases. These ratios can in principle be obtained by applying Particle Profiler to a single lipoprotein profile measurement, which makes clinical application feasible.