Benchmarking of mutation diagnostics in clinical lung cancer specimens

Treatment of EGFR-mutant non-small cell lung cancer patients with the tyrosine kinase inhibitors erlotinib or gefitinib results in high response rates and prolonged progression-free survival. Despite the development of sensitive mutation detection approaches, a thorough validation of these in a clinical setting has so far been lacking. We performed, in a clinical setting, a systematic validation of dideoxy ‘Sanger’ sequencing and pyrosequencing against massively parallel sequencing as one of the most sensitive mutation detection technologies available. Mutational annotation of clinical lung tumor samples revealed that of all patients with a confirmed response to EGFR inhibition, only massively parallel sequencing detected all relevant mutations. By contrast, dideoxy sequencing missed four responders and pyrosequencing missed two responders, indicating a dramatic lack of sensitivity of dideoxy sequencing, which is widely applied for this purpose. Furthermore, precise quantification of mutant alleles revealed a low correlation (r² = 0.27) of histopathological estimates of tumor content and frequency of mutant alleles, thereby questioning the use of histopathology for stratification of specimens for individual analytical procedures. Our results suggest that enhanced analytical sensitivity is critically required to correctly identify patients responding to EGFR inhibition. More broadly, our results emphasize the need for thorough evaluation of all mutation detection approaches against massively parallel sequencing as a prerequisite for any clinical implementation.     

Deriving biomedical diagnostics from NMR spectroscopic data

Biomedical spectroscopic experiments generate large volumes of data. For accurate, robust diagnostic tools the data must be analyzed for only a few characteristic observations per subject, and a large number of subjects must be studied. We describe here two of the current data analytic approaches applied to this problem: SIMCA (principal component analysis, partial least squares), and the statistical classification strategy (SCS). We demonstrate the application of the SCS by three examples of its use in analyzing ¹H NMR spectra: screening for colon cancer, characterization of thyroid cancer, and distinguishing cancer from cholangitis in the biliary tract.     

Benchmarking a memetic algorithm for ordering microarray data

This work introduces a new algorithm for "gene ordering". Given a matrix of gene expression data values, the task is to find a permutation of the gene names list such that genes with similar expression patterns should be relatively close in the permutation. The algorithm is based on a combined approach that integrates a constructive heuristic with evolutionary and Tabu Search techniques in a single methodology. To evaluate the benefits of this method, we compared our results with the current outputs provided by several widely used algorithms in functional genomics. We also compared the results with our own hierarchical clustering method when used in isolation. We show that the use of images, corrupted with known levels of noise, helps to illustrate some aspects of the performance of the algorithms and provide a complementary benchmark for the analysis. The use of these images, with known high-quality solutions, facilitates in some cases the assessment of the methods and helps the software development, validation and reproducibility of results. We also propose two quantitative measures of performance for gene ordering. Using these measures, we make a comparison with probably the most used algorithm (due to Eisen and collaborators, PNAS 1998) using a microarray dataset available on the public domain (the complete yeast cell cycle dataset).     

Pairwise dependence diagnostics for clustered failure-time data

Frailty and copula models specify a parametric dependence structurefor multivariate failure-time data. Estimation of some jointquantities can be highly sensitive to the assumed parametricform, and hence model fit is an important issue. This paperlays out a general diagnostic framework for evaluating and selectingfrailty and copula models. The approach is based on the cumulativesum of residuals that are calculated in bivariate time. Theresiduals reflect the difference between the observed and expectedbivariate association structures. The proposed model-checkingprocess is interpretable with a limiting distribution whichcan be approximated using the bootstrap. Simulations and a dataexample illustrate the practical application of the method.     

Validation data for periprosthetic bone remodelling theories

Periprosthetic adaptive bone remodelling after total hip arthroplasty (THA) has been frequently simulated in computer models, combining bone remodelling theory with finite element analysis. Unfortunately, there still subsist a lack of clinical data, which are necessary for validation of these simulation results. Therefore, the objective of the current project is to collect prospective volumetric bone density data with a clinical computerized tomography study in seven patients after THA. A retrospective study 12 years after implantation in 11 patients was added. A data set of about 100 000 bone voxels for each femur was collected. In all prospective cases, the predominant change is seen during the first year. The average density reduction in the horizontal slices was between 50 and 150 Hounsfield units (HU) (approx. 10%; p<0.001) after 2 years. Loss of density is particularly strong distal of the minor trochanter and decreases from proximal to distal.

For the 12 years retrospective study, the contralateral femur provided the control. Similar trends comparable to the prospective 2-year follow-up CT density values were seen in most cases with density reductions of up to 400 HU (30%). However, in one of these cases there was no difference between the operated and the control density.

As far as we are aware, this is the first collection of fully prospective 3D validation data in vivo for periprosthetic adaptive bone remodelling theories. The data are also unique as they are suitable for direct patient-specific 3D finite element meshing and individual weight-related loading.     

A geostatistical framework for combining spatially referenced disease prevalence data from multiple diagnostics

Multiple diagnostic tests are often used due to limited resources or because they provide complementary information on the epidemiology of a disease under investigation. Existing statistical methods to combine prevalence data from multiple diagnostics ignore the potential overdispersion induced by the spatial correlations in the data. To address this issue, we develop a geostatistical framework that allows for joint modelling of data from multiple diagnostics by considering two main classes of inferential problems: (a) to predict prevalence for a gold-standard diagnostic using low-cost and potentially biased alternative tests; (b) to carry out joint prediction of prevalence from multiple tests. We apply the proposed framework to two case studies: mapping Loa loa prevalence in Central and West Africa, using miscroscopy, and a questionnaire-based test called RAPLOA; mapping Plasmodium falciparum malaria prevalence in the highlands of Western Kenya using polymerase chain reaction and a rapid diagnostic test. We also develop a Monte Carlo procedure based on the variogram in order to identify parsimonious geostatistical models that are compatible with the data. Our study highlights (a) the importance of accounting for diagnostic-specific residual spatial variation and (b) the benefits accrued from joint geostatistical modelling so as to deliver more reliable and precise inferences on disease prevalence.     

Adequate selection of antigens for serologic diagnostics of experimental melioidosis

Sera of Burkholderia pseudomallei-infected golden hamsters, white mice, guinea pigs and white rats were studied. Immunochemical analysis revealed presence of antibodies against antigens 2, 3, 6, d, and g with significant predominance of antibodies to antigens 6 and d. Antigen d was detected irrespectively to strain used for experimental infection and experimental animal species. Antibodies to antigen 6 were detected only when strains belonging to Asian serovar, but not to Australian serovar, were used. On the basis of antigens 6 and d following methods of serologic diagnostics were developed: reaction of indirect hemagglutination, reaction of latex agglutination, and radioimmunological assay. Their sensitivity and specificity reached 100% during experimental melioidosis in golden hamsters, guinea pigs and white rats.     

Evaluation of alternate harvesting strategies using experimental microcosms

Experimental evidence to evaluate alternate conservation policies for harvested populations is currently meager. We used populations of the ciliate Tetrahymena thermophila growing in test tube microcosms to experimentally evaluate the effects of alternate harvesting policies in a controlled, replicable setting. Simple density-dependent models were effective in predicting patterns of ciliate population growth in the microcosms. We evaluated several univariate models, finding that a Ricker logistic model was a better predictor of ciliate population dynamics than Gompertz logistic, non-linear logistic, or random walk models. Using the Ricker logistic model as a demographic skeleton, we modeled ciliate population dynamics with respect to three alternate harvesting policies (fixed quota, fixed proportion, and fixed escapement), each conducted at four comparable levels of harvest intensity. The parameterized demographic models predicted that fixed quota harvesting would lead to lower mean ciliate abundance and higher temporal variability in ciliate abundance than fixed proportion or fixed escapement policies, with an appreciable risk of extinction, even under the controlled environmental conditions of our experimental system. For each harvesting policy, the intensity of harvest had demonstrable effects on population density. Population variability was higher for fixed quota harvesting than the other policies. The stochastic demographic model successfully predicted heightened extinction risk in the fixed quota system, relative to the other management treatments. Our experimental evidence lends support to the theoretical prediction that fixed quota harvesting is riskier than fixed proportion or fixed escapement policies.     

Instantaneous helical axis estimation from 3-D video data in neck kinematics for whiplash diagnostics

To date, the diagnosis of whiplash injuries has been very difficult and largely based on subjective, clinical assessment. The work by Winters and Peles Multiple Muscle Systems—Biomechanics and Movement Organization. Springer, New York (1990) suggests that the use of finite helical axes (FHAs) in the neck may provide an objective assessment tool for neck mobility. Thus, the position of the FHA describing head-trunk motion may allow discrimination between normal and pathological cases such as decreased mobility in particular cervical joints. For noisy, unsmoothed data, the FHAs must be taken over rather large angular intervals if the FHAs are to be reconstructed with sufficient accuracy; in the Winters and Peles study, these intervals were approximately 10°.

In order to study the movements' microstructure, the present investigation uses instantaneous helical axes (IHAs) estimated from low-pass smoothed video data. Here, the small-step noise sensitivity of the FHA no longer applies, and proper low-pass filtering allows estimation of the IHA even for small rotation velocity ω of the moving neck. For marker clusters mounted on the head and trunk, technical system validation showed that the IHAs direction dispersions were on the order of one degree, while their position dispersions were on the order of 1 mm, for low-pass cut-off frequencies of a few Hz (the dispersions were calculated from ω-weighted errors, in order to account for the adverse effects of vanishing ω).

Various simple, planar models relating the instantaneous 2-D centre of rotation with the geometry and kinematics of a multi-joint neck model are derived, in order to gauge the utility of the FHA and IHa approaches.

Some preliminary results on asymptomatic and pathological subjects are provided, in terms of the ‘ruled surface’ formed by sampled IHAs and of their piercing points through the mid-sagittal plane during a prescribed flexion-extension movement of the neck.     

Evidence for and against the existence of alternate attractors on coral reefs

Synthesis Coral reefs are widely thought to exhibit multiple attractors which have profound implications for people that depend on them. If reefs become ‘stuck’ within a self‐reinforcing state dominated by seaweed, it becomes disproportionately difficult and expensive for managers to shift the system back towards its natural, productive, coral state. The existence of multiple attractors is controversial. We assess various forms of evidence and conclude that there remains no incontrovertible proof of multiple attractors on reefs. However, the most compelling evidence, which combines ecological models and field data, is far more consistent with multiple attractors than the competing hypothesis of only a single, coral attractor. Managers should exercise caution and assume that degraded reefs can become stuck there. Testing for the existence of alternate attractors in ecosystems that possess slow dynamics and frequent pulse perturbation is exceptionally challenging. Coral reefs typify such conditions and the existence of alternate attractors is controversial. We analyse different forms of evidence and assess whether they support or challenge the existence of multiple attractors on Caribbean reefs, many of which have shown profound phase shifts in community structure from coral to algal dominance. Field studies alone provide no insight into multiple attractors because the non‐equilibrial nature of reef dynamics prevents equilibria from being observed. Statistical models risk failing to sample the parameter space in which multiple attractors occur, and have failed to account for the confounding effects of heterogeneous environments, anthropogenic drivers (e.g. fishing), and major disturbances (e.g. hurricanes). Simple and complex models all find multiple attractors over some – though not all – regions of a system driver (fishing). Tests of model predictions with field data closely match theory of alternate attractors but a forward‐leaning monotonic curve with only a single attractor can also be fitted to these data. Deeper consideration of the assumptions of this monotonic relationship reveal significant ecological problems which disappear under a model of multiple attractors. To date, there is no evidence against the existence of multiple attractors on Caribbean reefs and while there remains no definitive proof, the balance of evidence and ecological reasoning favours their existence. Theory predicts that Caribbean reefs do not exhibit alternate attractors in their natural state but that disease‐induced loss of two key functional groups has generated bistability. Whether alternate attractors becomes a persistent element of reef dynamics or a brief moment in their geological history will depend, in part, on the ability of functional groups to recover and the impacts of climate change and ocean acidification on coral growth and mortality.     

A method for determining constants of first-order reactions from experimental data

The authors have utilized a previously proposed mathematical equation (introduced originally for development of empirical equations) as a useful tool for evaluation of first-order reaction rate constants. By assigning physical significance to the parameter α, the equation can be utilized in obtaining excellent estimates for limiting boundary values and velocity constants.     

Reconstitution of the transition matrix from experimental data

The identification of a linear compartment model, which may describe a chemical or biological process, is a difficult task, since the available data is generally limited. In this paper we propose a method for determining the state transition matrix by minimizing a given quadratic criterion. To solve the resulting matrix equation, an assumption has to be made which constitutes a necessary condition for the identifiability of the model. Moreover when this assumption is satisfied, it is shown that the knowledge of one line or one column of the transition matrix is sufficient to define it completely.     

Automated method for modeling seven-helix transmembrane receptors from experimental data.

A rule-based automated method is presented for modeling the structures of the seven transmembrane helices of G-protein-coupled receptors. The structures are generated by using a simulated annealing Monte Carlo procedure that positions and orients rigid helices to satisfy structural restraints. The restraints are derived from analysis of experimental information from biophysical studies on native and mutant proteins, from analysis of the sequences of related proteins, and from theoretical considerations of protein structure. Calculations are presented for two systems. The method was validated through calculations using appropriate experimental information for bacteriorhodopsin, which produced a model structure with a root mean square (rms) deviation of 1.87 A from the structure determined by electron microscopy. Calculations are also presented using experimental and theoretical information available for bovine rhodopsin to assign the helices to a projection density map and to produce a model of bovine rhodopsin that can be used as a template for modeling other G-protein-coupled receptors.     

Scientific data and theories for salmonellosis dose–response assessment

An “expansive” risk assessment approach is illustrated, characterizing dose–response relationships for salmonellosis in light of the full body of evidence for human and murine superorganisms. Risk assessments often require analysis of costs and benefits for supporting public health decisions. Decision-makers and the public need to understand uncertainty in such analyses for two reasons. Uncertainty analyses provide a range of possibilities within a framework of present scientific knowledge, thus helping to avoid undesirable consequences associated with the selected policies. And, it encourages the risk assessors to scrutinize all available data and models, thus helping avoid subjective or systematic errors. Without the full analysis of uncertainty, decisions could be biased by judgments based solely on default assumptions, beliefs, and statistical analyses of selected correlative data. Alternative data and theories that incorporate variability and heterogeneity for the human and murine superorganisms, particularly colonization resistance, are emerging as major influences for microbial risk assessment. Salmonellosis risk assessments are often based on conservative default models derived from selected sets of outbreak data that overestimate illness. Consequently, the full extent of uncertainty of estimates of annual number of illnesses is not incorporated in risk assessments and the presently used models may be incorrect.     

Determining metabolic sensitivity coefficients directly from experimental data

The application of metabolic control theory (MCT), or other methods of determining metabolic sensitivity to the rates of specific cellular processes, such as enzymatic reactions, requires knowledge of the elasticity coefficients (system partial derivatives) for the processes under study. Although rate equations are available in the literature for some enzymatic reactions, there are many reactions and processes for which this is not the case. Although one could perform the experiments necessary to determine the rate equations for a given system, these equations are, in fact, not required for the calculation of sensitivities-only the elasticities (the derivatives) are needed. A more direct and efficient approach would be to compute elasticities directly from experimental data. Errors can analysis and alternative regression techniques are presented which not only allow one to eliminate data components with excessive noise, but also provide guidance as to what additional data may be require for accurate sensitivity analysis. This information indicates which measurements require more accuracy and what additional experiments should be conducted to reduce errors in calculated metabolic sensitivity coefficients. (c) 1993 Wiley & Sons, Inc.     

Iterative algorithms for processing experimental data

The need to solve linear and nonlinear integral equations arise, e.g., in recovering plasma parameters from the data of multichannel diagnostics. The paper presents an iterative method for solving integral equations with a singularity at the upper limit of integration. The method consists in constructing successive approximations and calculating the integral by quadrature formulas in each integration interval. An example of application of the iterative algorithm to numerically solve an integral equation similar to those arising in recovering the plasma density profile from reflectometry data is presented.     

Derivation of antibody distribution from experimental binding data.

It can be demonstrated that antibody populations cannot be suitably described in terms of a Sips distribution. Use of the Stieltjes transform allows instead derivation, from experimental binding data, of the most probable distribution of the association constants. From graphical interpolation of the experimental data four parameters can be obtained, which characterize a satisfactory bimodal distribution. By this procedure, analysis of data taken at different times of a humoral immune response, indicates that the relative abundance of the two sub-populations, rather than their mean association constant, is mainly affected by time or by variations of antigen dose. By use of the same procedure it is also possible to show that, at least as far as haptens are concerned, the slope of the 50% antigen-binding plot, often taken as a measure of the “avidity” of antibodies, is instead a function of the relative weight of the two sub-populations and of the spread of each of them.     

Fitting experimental data to models that use morphological data from public databases

Ideally detailed neuron models should make use of morphological and electrophysiological data from the same cell. However, this rarely happens. Typically a modeler will choose a cell morphology from a public database, assign standard values for R _a, C _m, and other parameters and then do the modeling study. The assumption is that the model will produce results representative of what might be obtained experimentally. To test this assumption we developed models of CA1 hippocampal pyramidal neurons using 4 different morphologies obtained from 3 public databases. The multiple run fitter in NEURON was used to fit parameter values in each of the 4 morphological models to match experimental data recorded from 19 CA1 pyramidal cells. Fits with fixed standard parameter values produced results that were generally not representative of our experimental data. However, when parameter values were allowed to vary, excellent fits were obtained in almost all cases, but the fitted parameter values were very different among the 4 reconstructions and did not match standard values. The differences in fitted values can be explained by very different diameters, total lengths, membrane areas and volumes among the reconstructed cells, reflecting either cell heterogeneity or issues with the reconstruction data. The fitted values compensated for these differences to make the database cells and experimental cells more similar electrotonically. We conclude that models using fully reconstructed morphologies need to be calibrated with experimental data (even when morphological and electrophysiological data come from the same cell), model results should be generated with multiple reconstructions, morphological and experimental cells should come from the same strain of animal at the same age, and blind use of standard parameter values in models that use reconstruction data may not produce representative experimental results. Action Editor: Steve Redman