Modelling High-Dimensional Data by Mixtures of Factor Analyzers

We focus on mixtures of factor analyzers from the perspective of a method for model-based density estimation from high-dimensional data, and hence for the clustering of such data. This approach enables a normal mixture model to be fitted to a sample of n data points of dimension p, where p is large relative to n. The number of free parameters is controlled through the dimension of the latent factor space. By working in this reduced space, it allows a model for each component-covariance matrix with complexity lying between that of the isotropic and full covariance structure models. We shall illustrate the use of mixtures of factor analyzers in a practical example that considers the clustering of cell lines on the basis of gene expressions from microarray experiments. (C) 2002 Elsevier Science B.V. All rights reserved.

Fitting genetic models to twin data with binary and ordered categorical responses: A comparison of structural equation modelling and Bayesian hierarchical models

We compare Bayesian methodology utilizing free-ware BUGS (Bayesian Inference Using Gibbs Sampling) with the traditional structural equation modelling approach based on another free-ware package, Mx. Dichotomous and ordinal (three category) twin data were simulated according to different additive genetic and common environment models for phenotypic variation. Practical issues are discussed in using Gibbs sampling as implemented by BUGS to fit subject-specific Bayesian generalized linear models, where the components of variation may be estimated directly. The simulation study (based on 2000 twin pairs) indicated that there is a consistent advantage in using the Bayesian method to detect a correct model under certain specifications of additive genetics and common environmental effects. For binary data, both methods had difficulty in detecting the correct model when the additive genetic effect was low (between 10 and 20%) or of moderate range (between 20 and 40%). Furthermore, neither method could adequately detect a correct model that included a modest common environmental effect (20%) even when the additive genetic effect was large (50%). Power was significantly improved with ordinal data for most scenarios, except for the case of low heritability under a true ACE model. We illustrate and compare both methods using data from 1239 twin pairs over the age of 50 years, who were registered with the Australian National Health and Medical Research Council Twin Registry (ATR) and presented symptoms associated with osteoarthritis occurring in joints of the hand.

Predicting the HER2 status of breast cancer from basic histopathology data: an analysis of 1500 breast cancers as part of the HER2000 International Study

The tests that are currently available for the measurement of overexpression of the human epidermal growth factor-2 (HER2) in breast cancer have shown considerable problems in accuracy and interlaboratory reproducibility. Although these problems are partly alleviated by the use of validated, standardised 'kits', there may be considerable cost involved in their use. Prior to testing it may therefore be an advantage to be able to predict from basic pathology data whether a cancer is likely to overexpress HER2. In this study, we have correlated pathology features of cancers with the frequency of HER2 overexpression assessed by immunohistochemistry (IHC) using HercepTest (Dako). In addition, fluorescence in situ hybridisation (FISH) has been used to re-test the equivocal cancers and interobserver variation in assessing HER2 overexpression has been examined by a slide circulation scheme. Of the 1536 cancers, 1144 (74.5%) did not overexpress HER2. Unequivocal overexpression (3+ by IHC) was seen in 186 cancers (12%) and an equivocal result (2+ by IHC) was seen in 206 cancers (13%). Of the 156 IHC 3+ cancers for which complete data was available, 149 (95.5%) were ductal NST and 152 (97%) were histological grade 2 or 3. Only 1 of 124 infiltrating lobular carcinomas (0.8%) showed HER2 overexpression. None of the 49 'special types' of carcinoma showed HER2 overexpression. Re-testing by FISH of a proportion of the IHC 2+ cancers showed that only 25 (23%) of those assessable exhibited HER2 gene amplification, but 46 of the 47 IHC 3+ cancers (98%) were confirmed as showing gene amplification. Circulating slides for the assessment of HER2 score showed a moderate level of agreement between pathologists (kappa 0.4). As a result of this study we would advocate consideration of a triage approach to HER-2 testing. Infiltrating lobular and special types of carcinoma may not need to be routinely tested at presentation nor may grade 1 NST carcinomas in which only 1.4% have been shown to overexpress HER2. Testing of these carcinomas may be performed when HER2 status is required to assist in therapeutic or other clinical/prognostic decision-making. The highest yield of HER2 overexpressing carcinomas is seen in the grade 3 NST subgroup in which 24% are positive by IHC. (C) 2003 Elsevier Science Ltd. All rights reserved.

Capillaries within compartments: Microvascular interpretation of dynamic positron emission tomography data

Measurement of exchange of substances between blood and tissue has been a long-lasting challenge to physiologists, and considerable theoretical and experimental accomplishments were achieved before the development of the positron emission tomography (PET). Today, when modeling data from modern PET scanners, little use is made of earlier microvascular research in the compartmental models, which have become the standard model by which the vast majority of dynamic PET data are analysed. However, modern PET scanners provide data with a sufficient temporal resolution and good counting statistics to allow estimation of parameters in models with more physiological realism. We explore the standard compartmental model and find that incorporation of blood flow leads to paradoxes, such as kinetic rate constants being time-dependent, and tracers being cleared from a capillary faster than they can be supplied by blood flow. The inability of the standard model to incorporate blood flow consequently raises a need for models that include more physiology, and we develop microvascular models which remove the inconsistencies. The microvascular models can be regarded as a revision of the input function. Whereas the standard model uses the organ inlet concentration as the concentration throughout the vascular compartment, we consider models that make use of spatial averaging of the concentrations in the capillary volume, which is what the PET scanner actually registers. The microvascular models are developed for both single- and multi-capillary systems and include effects of non-exchanging vessels. They are suitable for analysing dynamic PET data from any capillary bed using either intravascular or diffusible tracers, in terms of physiological parameters which include regional blood flow. (C) 2003 Elsevier Ltd. All rights reserved.

An EM-based Semi-Parametric Mixture Model Approach to the Regression Analysis of Competing-Risks Data

We consider a mixture model approach to the regression analysis of competing-risks data. Attention is focused on inference concerning the effects of factors on both the probability of occurrence and the hazard rate conditional on each of the failure types. These two quantities are specified in the mixture model using the logistic model and the proportional hazards model, respectively. We propose a semi-parametric mixture method to estimate the logistic and regression coefficients jointly, whereby the component-baseline hazard functions are completely unspecified. Estimation is based on maximum likelihood on the basis of the full likelihood, implemented via an expectation-conditional maximization (ECM) algorithm. Simulation studies are performed to compare the performance of the proposed semi-parametric method with a fully parametric mixture approach. The results show that when the component-baseline hazard is monotonic increasing, the semi-parametric and fully parametric mixture approaches are comparable for mildly and moderately censored samples. When the component-baseline hazard is not monotonic increasing, the semi-parametric method consistently provides less biased estimates than a fully parametric approach and is comparable in efficiency in the estimation of the parameters for all levels of censoring. The methods are illustrated using a real data set of prostate cancer patients treated with different dosages of the drug diethylstilbestrol. Copyright (C) 2003 John Wiley Sons, Ltd.

Effect of changing data collection parameters on statistical motor unit number estimates

The effect of number of samples and selection of data for analysis on the calculation of surface motor unit potential (SMUP) size in the statistical method of motor unit number estimates (MUNE) was determined in 10 normal subjects and 10 with amyotrophic lateral sclerosis (ALS). We recorded 500 sequential compound muscle action potentials (CMAPs) at three different stable stimulus intensities (10–50% of maximal CMAP). Estimated mean SMUP sizes were calculated using Poisson statistical assumptions from the variance of 500 sequential CMAP obtained at each stimulus intensity. The results with the 500 data points were compared with smaller subsets from the same data set. The results using a range of 50–80% of the 500 data points were compared with the full 500. The effect of restricting analysis to data between 5–20% of the CMAP and to standard deviation limits was also assessed. No differences in mean SMUP size were found with stimulus intensity or use of different ranges of data. Consistency was improved with a greater sample number. Data within 5% of CMAP size gave both increased consistency and reduced mean SMUP size in many subjects, but excluded valid responses present at that stimulus intensity. These changes were more prominent in ALS patients in whom the presence of isolated SMUP responses was a striking difference from normal subjects. Noise, spurious data, and large SMUP limited the Poisson assumptions. When these factors are considered, consistent statistical MUNE can be calculated from a continuous sequence of data points. A 2 to 2.5 SD or 10% window are reasonable methods of limiting data for analysis. Muscle Nerve 27: 320–331, 2003

Model-based clustering in gene expression microarrays: An application to breast cancer data

In microarray studies, the application of clustering techniques is often used to derive meaningful insights into the data. In the past, hierarchical methods have been the primary clustering tool employed to perform this task. The hierarchical algorithms have been mainly applied heuristically to these cluster analysis problems. Further, a major limitation of these methods is their inability to determine the number of clusters. Thus there is a need for a model-based approach to these. clustering problems. To this end, McLachlan et al. [7] developed a mixture model-based algorithm (EMMIX-GENE) for the clustering of tissue samples. To further investigate the EMMIX-GENE procedure as a model-based -approach, we present a case study involving the application of EMMIX-GENE to the breast cancer data as studied recently in van 't Veer et al. [10]. Our analysis considers the problem of clustering the tissue samples on the basis of the genes which is a non-standard problem because the number of genes greatly exceed the number of tissue samples. We demonstrate how EMMIX-GENE can be useful in reducing the initial set of genes down to a more computationally manageable size. The results from this analysis also emphasise the difficulty associated with the task of separating two tissue groups on the basis of a particular subset of genes. These results also shed light on why supervised methods have such a high misallocation error rate for the breast cancer data.

Zygosity Diagnosis in the Absence of Genotypic Data: An Approach Using Latent Class Analysis

For zygosity diagnosis in the absence of genotypic data, or in the recruitment phase of a twin study where only single twins from same-sex pairs are being screened, or to provide a test for sample duplication leading to the false identification of a dizygotic pair as monozygotic, the appropriate analysis of respondents' answers to questions about zygosity is critical. Using data from a young adult Australian twin cohort (N = 2094 complete pairs and 519 singleton twins from same-sex pairs with complete responses to all zygosity items), we show that application of latent class analysis (LCA), fitting a 2-class model, yields results that show good concordance with traditional methods of zygosity diagnosis, but with certain important advantages. These include the ability, in many cases, to assign zygosity with specified probability on the basis of responses of a single informant (advantageous when one zygosity type is being oversampled); and the ability to quantify the probability of misassignment of zygosity, allowing prioritization of cases for genotyping as well as identification of cases of probable laboratory error. Out of 242 twins (from 121 like-sex pairs) where genotypic data were available for zygosity confirmation, only a single case was identified of incorrect zygosity assignment by the latent class algorithm. Zygosity assignment for that single case was identified by the LCA as uncertain (probability of being a monozygotic twin only 76%), and the co-twin's responses clearly identified the pair as dizygotic (probability of being dizygotic 100%). In the absence of genotypic data, or as a safeguard against sample duplication, application of LCA for zygosity assignment or confirmation is strongly recommended.

Systematic characterization of mutations in yeast acetohydroxyacid synthase: Interpretation of herbicide-resistance data

Acetohydroxyacid synthase (AHAS, EC 4.1.3.18) catalyses the first step in branched-chain amino acid biosynthesis and is the target for sulfonylurea and imidazolinone herbicides, which act as potent and specific inhibitors. Mutants of the enzyme have been identified that are resistant to particular herbicides. However, the selectivity of these mutants towards various sulfonylureas and imidazolinones has not been determined systematically. Now that the structure of the yeast enzyme is known, both in the absence and presence of a bound herbicide, a detailed understanding of the molecular interactions between the enzyme and its inhibitors becomes possible. Here we construct 10 active mutants of yeast AHAS, purify the enzymes and determine their sensitivity to six sulfonylureas and three imidazolinones. An additional three active mutants were constructed with a view to increasing imidazolinone sensitivity. These three variants were purified and tested for their sensitivity to the imidazolinones only. Substantial differences are observed in the sensitivity of the 13 mutants to the various inhibitors and these differences are interpreted in terms of the structure of the herbicide-binding site on the enzyme.

Linking physiological and architectural models of cotton

Despite the strong influence of plant architecture on crop yield, most crop models either ignore it or deal with it in a very rudimentary way. This paper demonstrates the feasibility of linking a model that simulates the morphogenesis and resultant architecture of individual cotton plants with a crop model that simulates the effects of environmental factors on critical physiological processes and resulting yield in cotton. First the varietal parameters of the models were made concordant. Then routines were developed to allocate the flower buds produced each day by the crop model amongst the potential positions generated by the architectural model. This allocation is done according to a set of heuristic rules. The final weight of individual bolls and the shedding of buds and fruit caused by water, N, and C stresses are processed in a similar manner. Observations of the positions of harvestable fruits, both within and between plants, made under a variety of agronomic conditions that had resulted in a broad range of plant architectures were compared to those predicted by the model with the same environmental inputs. As illustrated by comparisons of plant maps, the linked models performed reasonably well, though performance of the fruiting point allocation and shedding algorithms could probably be improved by further analysis of the spatial relationships of retained fruit. (C) 2002 Elsevier Science Ltd. All rights reserved.

Evaluating a model of parental influence on youth physical activity

Objective: To test a conceptual model linking parental physical activity orientations, parental support for physical activity, and children's self-efficacy perceptions with physical activity participation. Participants and Setting: The sample consisted of 380 students in grades 7 through 12 (mean age, 14.0 +/- 1.6 years) and their parents. Data collection took place during the fall of 1996. Main Outcome Measures: Parents completed a questionnaire assessing their physical activity habits, enjoyment of physical activity, beliefs regarding the importance of physical activity, and supportive behaviors for their child's physical activity. Students completed a 46-item inventory assessing physical activity during the previous 7 days and a 5-item physical activity self-efficacy scale. The model was tested via observed variable path analysis using structural equation modeling techniques (AMOS 4.0). Results: An initial model, in which parent physical activity orientations predicted child physical activity via parental support and child self-efficacy, did not provide an acceptable fit to the data. Inclusion of a direct path from parental support to child physical activity and deletion of a nonsignificant path from parental physical activity to child physical activity significantly improved model fit. Standardized path coefficients for the revised model ranged from 0.17 to 0.24, and all were significant at the p < 0.0001 level. Conclusions: Parental support was an important correlate of youth physical activity, acting directly or indirectly through its influence on self-efficacy. Physical activity interventions targeted at youth should include and evaluate the efficacy of individual-level and community-level strategies to increase parents' capacity to provide instrumental and motivational support for their children's physical activity.

G-values for scission and crosslinking on γ-radiolysis of Ultem at 303 K

The radiation chemical yields G(S) and G(X) for H-linking and Y-linking models for Ultem have been calculated from molecular weight analysis by gel permeation chromatography. These G-values have been compared with the G-values obtained from analysis of soluble fractions above the gel dose, Which have been reported in previous works. An analysis of the molecular weight data in terms of H-linking and Y-linking mechanisms yielded values of G(S-H) = 1.0 x 10(-3) and G(H) = 6.0 x 10(-3) and G(S-Y) = 1.3 x 10(-2) and G(Y) = 1.8 x 10(-2). The corresponding values obtained from the solubility data were G(SH) = 0.53 x 10(-2), G(H) = 1.39 x 10(-2), G(S-Y) = 4.2 x 10(-2) and G(Y) - 4.6 x 10(-2). The origin of the disagreement between the molecular weight and solubility values is not clear, but it could arise as a result of observed microgel formation below the reported gel dose of 0.13 MGy. Whether the crosslink mechanism proceeds by an H-linking or Y-linking process is also unclear and will require direct observation of the crosslinking structures.

Cross-linking of PFA by electron beam irradiation

The effect of electron beam radiation on a perfluoroalkoxy (PFA) resin was examined using solid-state high-speed magic angle spinning F-19 NMR spectroscopy and FT-IR spectroscopy. Samples were prepared for analysis by subjecting them to electron beam radiation in the dose range 0.5-2.0 MGy at 633 K, which is above the crystalline melting temperature. The new structures were identified and include new saturated chain ends, short and long branches, unsaturated groups, and cross-links. The radiation chemical yield (G value) of new long branch points was greater than the G value of new chain ends, suggesting that cross-linking is the net radiolytic process. This conclusion was supported by an observed decrease in the crystallinity and an increase in the optical clarity of the polymer.

Data Mart para apoio ?? capacita????o em gest??o de pol??ticas p??blicas voltado para uma Escola de Administra????o P??blica

Este trabalho consiste num estudo de caso que se destina ao desenvolvimento de um Data Mart que possibilite a Escola Nacional de Administra????o P??blica ??? ENAP conhecer o perfil e o panorama geral da situa????o funcional dos servidores p??blicos federais que se capacitaram na Escola nos ??ltimos 7 anos. O aplicativo foi desenvolvido cruzando o banco de dados do sistema gerenciador dos cursos ministrados pela ENAP, onde est??o armazenadas informa????es sobre os alunos capacitados, os cursos realizados, os resultados alcan??ados, o perfil dos docentes e demais informa????es relativas ??s atividades da Escola, com os dados gerados pelo Sistema Integrado de Administra????o de Recursos Humanos ??? SIAPE, cuja extra????o de dados foi direcionada para os registros sobre a situa????o funcional, cargos, carreiras, fun????es, ??rg??os e alguns dados pessoais dos alunos, servidores p??blicos federais que se encontram registrados no SIAPE