Congenital rubella infection is a significant cause of birth defects worldwide. In the UK, vaccination programmes, introduced initially in 1970, have been successful in preventing congenital rubella cases and interrupting virus circulation. Previous studies have shown that women with a non-British background are more likely to be susceptible to rubella in pregnancy. London, with a large migrant population and suboptimal vaccination uptake is at particular risk of outbreaks. This thesis presents results of a rubella serosurvey in pregnant women in the North Thames area, using 18882 newborn dried blood spot samples to estimate maternally derived rubella IgG antibody level as a proxy for maternal levels, and investigate maternal age, Strategic Health Authority of residence and country of birth as risk factors for rubella seronegative status. Data on maternal country of birth was available through linkage to birth registration data, and latent class mixture models were used to identify samples which were seronegative for rubella IgG. Results of a study to validate two commercial enzyme-linked immunosorbent assays for use on dried blood spot sample are also presented. Overall, 2.7% of pregnant women were estimated to be rubella seronegative (95% CI 2.4% to 3.0%). Non-UK born women were significantly more likely to be seronegative to rubella for women born in South Asia, Sub-Saharan Africa and the Americas the adjusted odds ratios of rubella seronegative status were 4.96, 4.19 and 3.42 (95% CI 3.76 to 6.52, 3.11 to 5.64 and 2.13 to 5.52 respectively). Women aged less than 20 years were significantly more likely to be seronegative than women in their thirties (adjusted odds ratio 2.53, 95% CI 1.72 to 3.72). Application of latent class mixture regression models to estimate the independent effect of maternal characteristics on rubella antibody level is also demonstrated. These results highlight the continued risk to migrant women of rubella infection in pregnancy.

No description available.

No description available.

This thesis aims to investigate immunological and virological markers of HIV infection during pregnancy and to inform understanding of the impact of antenatal antiretroviral therapy (ART) regimens on such markers, using data from the European Collaborative Study (ECS). Since 1986, over 7,000 HIV infected pregnant women have been enrolled from 23 centres in 10 countries. Maternal information collected included timing and type of treatment regimen, repeated measurements on maternal CD4 counts and viral load (since 1987) and socio-demographic variables. Repeated HIV RNA measurements over pregnancy in over 300 untreated women were examined using linear mixed effects (LME) models. Immunological (CD4) and virological changes over pregnancy in 162 women receiving highly-active ART (HAART) at conception and throughout pregnancy were examined with a piecewise LME model, using a conditional likelihood approach to account for left-censored measurements. A bivariate LME model was used to assess the correlation between the two markers. Viral response to initial HAART regimens in 240 ART-naive pregnant women was determined using interval censored regression, with a propensity score to reduce treatment allocation bias. A Cox regression model was used to assess the effect of HAART on risk of premature delivery, treating elective caesarean section deliveries as right-censored outcomes. HIV RNA viral load was estimated to decrease over the second and third trimesters of pregnancy in untreated and treated women. Differences in levels of HIV RNA viral load and CD4 count by race and history of injecting drug use were identified. The rate of achieving undetectable viral load was greater for women initiating on nevirapine- containing HAART than for women on protease inhibitor-based HAART. There was no evidence of an increased risk of birth defects or prematurity associated with type or timing of HAART. These findings contribute to the evidence-base for management and understanding of HIV infection in pregnancy.

This thesis examines the application of a particular neuromorphic computational model, the Boltzmann Machine, to the evaluation of laryngeal behaviour using parameters derived from the acoustic analysis of irregularities in the periodic structures of speech signals. Over the last twenty five years, researchers in various fields such as speech science, laryngology, speech pathology and phonetics have demonstrated a growing interest in the acoustic characterisation of healthy and pathological voices. This research activity has been in response to the need for non-invasive and quantitative techniques for the assessment of layrngeal function. Over the past five years neuromorphic computation has undergone a dramatic transformation with the development of powerful learning algorithms and the promise of highly parallel implementations taking advantage of developments in high density integrated circuit technology. These neuromorphic systems are machines that behave in brain-like ways and compute by absorbing experience. The Boltzmann Machine learning algorithm provides a formally guaranteed procedure for performing gradient descent in a global error measure. This thesis presents, for the first time, results which demonstrate the potential of the Boltzmann Machine approach to the detection of laryngeal pathologies. A simulation environment for Boltzmann Machines was successfully developed which provided acceptable speeds of operation for the sizes of network investigated, and the quantity of training data used, provided approximations to the <i>theoretical</i> Boltzmann Machine were made. Chapter 4 presents details of this implementation. Experiments using various topologies of Boltzmann Machine made use of ten intonation and perturbation parameters, derived from the analysis of waveform perturbations of fundamental frequency and amplitude evidenced in samples of connected speech from groups of healthy and pathological male speakers. A series of experiments are presented in Chapter 6 which evaluate the performance of various Boltzmann Machine topologies and data representation formats to the differentiation between groups of healthy speakers and speakers with known pathological conditions of the larynx. From these experiments it was concluded that the intonation and perturbation parameters could be processed using a Boltzmann Machine to provide useful differentiation between groups of healthy speakers and speakers with known pathological conditions. Chapter 7 presents a series of experiments using various topologies of Boltzmann Machine to differentiate between broad classes of pathologies using ten intonation and perturbation parameters. The experiments showed that it was possible for various pathology groups to be differentiated in a training group. The successful development of a neuromorphic system for determination of laryngeal function associated with healthy and pathological phonation has a number of potential applications, including screening, differential diagnosis and tracking changes in the conditions of laryngeal pathology.

Until recently, deterministic methods have been used to estimate food risk. These methods give one risk estimate for the whole population, mostly ignore the variability in the consumption data and produce unrealistic results. Recently there has been an increase in the use of probabilistic models for food risk assessment. Some of these probabilistic models are purely based on empirical distributions and others on probability models in a frequentist or Bayesian framework. In this thesis we improve on existing Bayesian models for food consumption data collected on successive days, and propose possible models to assess various types of food risk. Bayesian hierarchical modelling provides a natural framework for risk assessment, and allows us to account for the various sources of variability present in the data. We discuss general problems associated for the various sources of variability present in the data. We discuss general problems associated with dietary data such as large proportions of zeros and extreme intakes, and suggest models to account for these. We look at ways to model intake of several food products which may all contain the same pesticide, and continue this with pesticide residue data for exposure assessment for that pesticide. We also discuss a non-Bayesian approach to study extreme intakes using Extreme-value theory. We use our models to produce predicted probabilities of exceeding recommended and safe levels of consumption for individual days and longer periods.

New developments in computer-based (<i>‘in-silico’</i>) technologies, combined with the exponential growth of publicly available information on potential therapeutic targets, provide a timely opportunity to develop a descriptor-based, virtual screening platform to fast-track ligand selection. Edinburgh University Ligand Selection System, known as EDULISS, aims to meet the database mining needs of the academic community through establishing a large depository of molecular descriptor values relating to over 1.5 million freely available chemicals. Embedding information from over 1500 different topological geometric, physiochemical and toxicological molecular descriptors in a ‘query-able’ MySQL database, provides a method to establish an ADMET (absorption, distribution, metabolism, excretion, and toxicity) context to these otherwise obscure descriptors. EDULISS has been applied to screen 16 prominent supplier chemical catalogues, identifying correlations and cut-off limits between descriptors used in traditional rule-based screens (e.g. rule of five, lead-like & rule of three) and those used in OSAR studies, that show strong correlation to chemical properties mediating protein-ligand interactions. Newly developed substructure and similarity searches, which are reliant on molecular descriptor values stored in EDULISS, have identified a wider pool of tetraethyl ammonium (TEA) based components as potential inhibitors for the aquaporin family of orphan proteins. Combined use of EDULISS in conjunction with molecular docking packages has provided modelling results supporting experimental evidence that suggests an alternative binding position of larger molecular weight TEA analogue inhibitors of AQP1. EDULISS has shown to be a novel chemical informatics tool for the ‘bulk’ analysis of molecular descriptors associated with large compendiums of ‘in-silico’ chemicals with the potential for establishing new rule-based database mining techniques.

Aim and Objectives: To explore the feasibility of accessing biological information and associated health information through a standards-based electronic health record. The objectives include constructing: a condition specific knowledge base prototype; an EHR system prototype based on a standard record architecture; and an interface that connects the two. Method: An ontology was constructed to organise biological and health information in a formal and structured way. Cystic fibrosis was selected as an exemplar condition and the Continuity of Care Record was selected for an EHR prototype application. The sequence variations information and health information in the knowledge base are presented through the EHR prototype's interface and the results are evaluated. Results: A substantive knowledge base prototype of cystic fibrosis was constructed. The content includes: the most common genetic mutations related to cystic fibrosis; time-oriented descriptions of cystic fibrosis; Cochrane conclusions; and gene therapy for cystic fibrosis. The content is organised on both time and problem oriented axes. It was found to be possible to present bio-health information that was case-specific through the EHR prototype interface. Conclusion: Sequence variations information and associated health information can be made accessible through a standards-based electronic health record prototype. Complex knowledge can be accessed, to some extent automatically, thereby providing a starting point for integrating formal and structured biological information within health record systems which can be deployed in clinical settings.

This work examines patterns of under-nutrition prevalence, based on data gathered through a growth monitoring programme run as part of a health information system by the Zimbabwean government. Most of the thesis makes use of secondary data - collected by other investigators for different purposes - but use is also made of a questionnaire survey of households in the Buhera district of Zimbabwe. Patterns of attendance and bias are assessed initially through a comparison of health information system statistics with census and resulted in an independent, community-based anthropometric survey. Under-nutrition prevalence data from the growth monitoring scheme are examined using regression analysis to identify temporal and spatial patterns. The pre-harvest period between January and March is found to be the season when underweight prevalence is greatest. Levels of under-nutrition are also found to have declined during the information system's life-time, though this apparent trend may be related to changing patterns of attendance. Agricultural, meteorological, infra-structural, and socio-economic data are collated within a Geographic Information System and related to the observed patterns of under-nutrition prevalence. Spatially, the areas which suffer from persistent problems of under-nutrition are those where rural poverty is greatest. Yearly fluctuations in district-level under-nutrition are found to be related to changes in rainfall, whilst monthly fluctuations are correlated with reported diarrhoea cases among children. A spatial simulation model of growth monitoring attendance is constructed for Buhera district, using the household survey data. The model's performance is poor when compared to health facility statistics, but the methodology developed enables data from different sources and different scales to be integrated. It is concluded that the methodology developed for analysing growth monitoring data in the thesis may have application elsewhere in southern Africa.

The aim of this thesis is to understand the genetic basis of the variation of human quantitative traits using data from twins and (to some extent) their families. Traits investigated include behavioural traits (intelligence), clinical traits (the metabolic syndrome) and anthropometric measures (height). The importance of human twins for understanding genetic variation in human quantitative traits is reviewed. The use of a novel finite mixture distribution model to partition phenotypic co(variance) of a trait into underlying genetic and environmental factors from twins of unknown zygosity is presented. The Scottish Mental Surveys of 1^st June 1932 and 4^th June 1947, respectively, administered the same validated verbal reasoning test (the Moray House Test) to almost everyone born in 1921 or 1936 and attending school in Scotland. Information on zygosity was unavailable. A novel application of a finite mixture distribution model estimated a large and consistent heritability of cognitive ability of about ~0/7. This study is the first to estimate genetic and environmental components of cognitive ability in entire school-attending populations and implies that large (national) data collections can provide sufficient information on twin pairs to estimate genetic parameters, even without known zygosity. The precision and bias of the finite mixture distribution model were assessed using computer simulations and application to IQ measures from a large sample of known zygosity twins (i.e. twins from the UK Twins’ Early Development Studies). It is shown that the mixture distribution is unbiased provided that the twins’ trait values are (bivariate) normally distributed and the sample size is large. However, it the bivariate normality assumption is violated, then the mixture distribution provides biased estimates.  The extension of the model to multivariate analysis is discussed. The simulations show that multivariate analysis decreases the standard error of the variance component estimates. Another statistical model, a mixed linear model is used to partition the phenotypic (co)variances of traits into genetic and environmental factors from twins of known zygosity (twins from the Danish Twin Registry). Its application to understand underlying genetic and environmental aetiology showed that endophenotypes associated with the metabolic syndrome do not appear to share a substantial common genetic or familial environmental background. Finally, a genome-wide linkage analysis to identify gene/chromosomal regions associated with adult height reveals several chromosomal regions that showed a modest linkage to adult height. This analysis provides further evidence for the polygenic nature of body height.