A comparative review of dimension reduction methods in approximate Bayesian computation

Approximate Bayesian computation (ABC) methods make use of comparisons between simulated and observed summary statistics to overcome the problem of computationally intractable likelihood functions. As the practical implementation of ABC requires computations based on vectors of summary statistics, rather than full data sets, a central question is how to derive low-dimensional summary statistics from the observed data with minimal loss of information. In this article we provide a comprehensive review and comparison of the performance of the principal methods of dimension reduction proposed in the ABC literature. The methods are split into three nonmutually exclusive classes consisting of best subset selection methods, projection techniques and regularization. In addition, we introduce two new methods of dimension reduction. The first is a best subset selection method based on Akaike and Bayesian information criteria, and the second uses ridge regression as a regularization procedure. We illustrate the performance of these dimension reduction techniques through the analysis of three challenging models and data sets.

Data Assimilation and Inverse Methods in Terms of a Probabilistic Formulation

The weak-constraint inverse for nonlinear dynamical models is discussed and derived in terms of a probabilistic formulation. The well-known result that for Gaussian error statistics the minimum of the weak-constraint inverse is equal to the maximum-likelihood estimate is rederived. Then several methods based on ensemble statistics that can be used to find the smoother (as opposed to the filter) solution are introduced and compared to traditional methods. A strong point of the new methods is that they avoid the integration of adjoint equations, which is a complex task for real oceanographic or atmospheric applications. they also avoid iterative searches in a Hilbert space, and error estimates can be obtained without much additional computational effort. the feasibility of the new methods is illustrated in a two-layer quasigeostrophic model.

Structure determination from powder diffraction data

Advances made over the past decade in structure determination from powder diffraction data are reviewed with particular emphasis on algorithmic developments and the successes and limitations of the technique. While global optimization methods have been successful in the solution of molecular crystal structures, new methods are required to make the solution of inorganic crystal structures more routine. The use of complementary techniques such as NMR to assist structure solution is discussed and the potential for the combined use of X-ray and neutron diffraction data for structure verification is explored. Structures that have proved difficult to solve from powder diffraction data are reviewed and the limitations of structure determination from powder diffraction data are discussed. Furthermore, the prospects of solving small protein crystal structures over the next decade are assessed.

A practical approach to goal modelling for time-constrained projects

Goal modelling is a well known rigorous method for analysing problem rationale and developing requirements. Under the pressures typical of time-constrained projects its benefits are not accessible. This is because of the effort and time needed to create the graph and because reading the results can be difficult owing to the effects of crosscutting concerns. Here we introduce an adaptation of KAOS to meet the needs of rapid turn around and clarity. The main aim is to help the stakeholders gain an insight into the larger issues that might be overlooked if they make a premature start into implementation. The method emphasises the use of obstacles, accepts under-refined goals and has new methods for managing crosscutting concerns and strategic decision making. It is expected to be of value to agile as well as traditional processes.

Century-of-Information Research (CIR): a strategy for research and innovation in the century of information

More data will be produced in the next five years than in the entire history of human kind, a digital deluge that marks the beginning of the Century of Information. Through a year-long consultation with UK researchers, a coherent strategy has been developed, which will nurture Century-of-Information Research (CIR); it crystallises the ideas developed by the e-Science Directors' Forum Strategy Working Group. This paper is an abridged version of their latest report which can be found at: http://wikis.nesc.ac.uk/escienvoy/Century_of_Information_Research_Strategy which also records the consultation process and the affiliations of the authors. This document is derived from a paper presented at the Oxford e-Research Conference 2008 and takes into account suggestions made in the ensuing panel discussion. The goals of the CIR Strategy are to facilitate the growth of UK research and innovation that is data and computationally intensive and to develop a new culture of 'digital-systems judgement' that will equip research communities, businesses, government and society as a whole, with the skills essential to compete and prosper in the Century of Information. The CIR Strategy identifies a national requirement for a balanced programme of coordination, research, infrastructure, translational investment and education to empower UK researchers, industry, government and society. The Strategy is designed to deliver an environment which meets the needs of UK researchers so that they can respond agilely to challenges, can create knowledge and skills, and can lead new kinds of research. It is a call to action for those engaged in research, those providing data and computational facilities, those governing research and those shaping education policies. The ultimate aim is to help researchers strengthen the international competitiveness of the UK research base and increase its contribution to the economy. The objectives of the Strategy are to better enable UK researchers across all disciplines to contribute world-leading fundamental research; to accelerate the translation of research into practice; and to develop improved capabilities, facilities and context for research and innovation. It envisages a culture that is better able to grasp the opportunities provided by the growing wealth of digital information. Computing has, of course, already become a fundamental tool in all research disciplines. The UK e-Science programme (2001-06)—since emulated internationally—pioneered the invention and use of new research methods, and a new wave of innovations in digital-information technologies which have enabled them. The Strategy argues that the UK must now harness and leverage its own, plus the now global, investment in digital-information technology in order to spread the benefits as widely as possible in research, education, industry and government. Implementing the Strategy would deliver the computational infrastructure and its benefits as envisaged in the Science & Innovation Investment Framework 2004-2014 (July 2004), and in the reports developing those proposals. To achieve this, the Strategy proposes the following actions: support the continuous innovation of digital-information research methods; provide easily used, pervasive and sustained e-Infrastructure for all research; enlarge the productive research community which exploits the new methods efficiently; generate capacity, propagate knowledge and develop skills via new curricula; and develop coordination mechanisms to improve the opportunities for interdisciplinary research and to make digital-infrastructure provision more cost effective. To gain the best value for money strategic coordination is required across a broad spectrum of stakeholders. A coherent strategy is essential in order to establish and sustain the UK as an international leader of well-curated national data assets and computational infrastructure, which is expertly used to shape policy, support decisions, empower researchers and to roll out the results to the wider benefit of society. The value of data as a foundation for wellbeing and a sustainable society must be appreciated; national resources must be more wisely directed to the collection, curation, discovery, widening access, analysis and exploitation of these data. Every researcher must be able to draw on skills, tools and computational resources to develop insights, test hypotheses and translate inventions into productive use, or to extract knowledge in support of governmental decision making. This foundation plus the skills developed will launch significant advances in research, in business, in professional practice and in government with many consequent benefits for UK citizens. The Strategy presented here addresses these complex and interlocking requirements.

Marine ecosystem models for earth systems applications: the MarQUEST experience

The MarQUEST (Marine Biogeochemistry and Ecosystem Modelling Initiative in QUEST) project was established to develop improved descriptions of marine biogeochemistry, suited for the next generation of Earth system models. We review progress in these areas providing insight on the advances that have been made as well as identifying remaining key outstanding gaps for the development of the marine component of next generation Earth system models. The following issues are discussed and where appropriate results are presented; the choice of model structure, scaling processes from physiology to functional types, the ecosystem model sensitivity to changes in the physical environment, the role of the coastal ocean and new methods for the evaluation and comparison of ecosystem and biogeochemistry models. We make recommendations as to where future investment in marine ecosystem modelling should be focused, highlighting a generic software framework for model development, improved hydrodynamic models, and better parameterisation of new and existing models, reanalysis tools and ensemble simulations. The final challenge is to ensure that experimental/observational scientists are stakeholders in the models and vice versa.

The binding site distance test score: a robust method for the assessment of predicted protein binding sites

We propose a novel method for scoring the accuracy of protein binding site predictions – the Binding-site Distance Test (BDT) score. Recently, the Matthews Correlation Coefficient (MCC) has been used to evaluate binding site predictions, both by developers of new methods and by the assessors for the community wide prediction experiment – CASP8. Whilst being a rigorous scoring method, the MCC does not take into account the actual 3D location of the predicted residues from the observed binding site. Thus, an incorrectly predicted site that is nevertheless close to the observed binding site will obtain an identical score to the same number of nonbinding residues predicted at random. The MCC is somewhat affected by the subjectivity of determining observed binding residues and the ambiguity of choosing distance cutoffs. By contrast the BDT method produces continuous scores ranging between 0 and 1, relating to the distance between the predicted and observed residues. Residues predicted close to the binding site will score higher than those more distant, providing a better reflection of the true accuracy of predictions. The CASP8 function predictions were evaluated using both the MCC and BDT methods and the scores were compared. The BDT was found to strongly correlate with the MCC scores whilst also being less susceptible to the subjectivity of defining binding residues. We therefore suggest that this new simple score is a potentially more robust method for future evaluations of protein-ligand binding site predictions.

Transgenic crops: the current and next generations.

This chapter describes the present status and future prospects for transgenic (genetically modified) crops. It concentrates on the most recent data obtained from patent databases and field trial applications, as well as the usual scientific literature. By these means, it is possible to obtain a useful perspective into future commercial products and international trends. The various research areas are subdivided on the basis of those associated with input (agronomic) traits and those concerned with output (e.g., food quality) characteristics. Among the former group are new methods of improving stress resistance, and among the latter are many examples of producing pharmaceutical compounds in plants.

Opportunities for reduction in acute toxicity testing via improved design

The conventional method for assessing acute oral toxicity (OECD Test Guideline 401) was designed to identify the median lethal dose (LD50), using the death of animals as an endpoint. Introduced as an alternative method (OECD Test Guideline 420), the Fixed Dose Procedure (FDP) relies on the observation of clear signs of toxicity, uses fewer animals and causes less suffering. More recently, the Acute Toxic Class method and the Up-and-Down Procedure have also been adopted as OECD test guidelines. Both of these methods also use fewer animals than the conventional method, although they still use death as an endpoint. Each of the three new methods incorporates a sequential dosing procedure, which results in increased efficiency. In 1999, with a view to replacing OECD Test Guideline 401, the OECD requested that the three new test guidelines be updated. This was to bring them in line with the regulatory needs of all OECD Member Countries, provide further reductions in the number of animals used, and introduce refinements to reduce the pain and distress experienced by the animals. This paper describes a statistical modelling approach for the evaluation of acute oral toxicity tests, by using the revised FDP for illustration. Opportunities for further design improvements are discussed.

Meta-analysis: a unifying meta-likelihood approach framing unobserved heterogeneity, study covariates, publication bias, and study quality

OBJECTIVES: This contribution provides a unifying concept for meta-analysis integrating the handling of unobserved heterogeneity, study covariates, publication bias and study quality. It is important to consider these issues simultaneously to avoid the occurrence of artifacts, and a method for doing so is suggested here. METHODS: The approach is based upon the meta-likelihood in combination with a general linear nonparametric mixed model, which lays the ground for all inferential conclusions suggested here. RESULTS: The concept is illustrated at hand of a meta-analysis investigating the relationship of hormone replacement therapy and breast cancer. The phenomenon of interest has been investigated in many studies for a considerable time and different results were reported. In 1992 a meta-analysis by Sillero-Arenas et al. concluded a small, but significant overall effect of 1.06 on the relative risk scale. Using the meta-likelihood approach it is demonstrated here that this meta-analysis is due to considerable unobserved heterogeneity. Furthermore, it is shown that new methods are available to model this heterogeneity successfully. It is argued further to include available study covariates to explain this heterogeneity in the meta-analysis at hand. CONCLUSIONS: The topic of HRT and breast cancer has again very recently become an issue of public debate, when results of a large trial investigating the health effects of hormone replacement therapy were published indicating an increased risk for breast cancer (risk ratio of 1.26). Using an adequate regression model in the previously published meta-analysis an adjusted estimate of effect of 1.14 can be given which is considerably higher than the one published in the meta-analysis of Sillero-Arenas et al. In summary, it is hoped that the method suggested here contributes further to a good meta-analytic practice in public health and clinical disciplines.

Recent developments in genetic data analysis: what can they tell us about human demographic history?

Over the last decade, a number of new methods of population genetic analysis based on likelihood have been introduced. This review describes and explains the general statistical techniques that have recently been used, and discusses the underlying population genetic models. Experimental papers that use these methods to infer human demographic and phylogeographic history are reviewed. It appears that the use of likelihood has hitherto had little impact in the field of human population genetics, which is still primarily driven by more traditional approaches. However, with the current uncertainty about the effects of natural selection, population structure and ascertainment of single-nucleotide polymorphism markers, it is suggested that likelihood-based methods may have a greater impact in the future.

Aziridine synthesis via nucleophilic attack of carbene equivalents on imines: the aza-Darzens reaction

The development of new methods for the efficient synthesis of aziridines has been of considerable interest to researchers for more than 60 years, but no single method has yet emerged as uniformly applicable, especially for asymmetric synthesis of chiral aziridines. One method which has been intensely examined and expanded of late involves the nucleophilic addition to imines by anions bearing a-leaving groups; by analogy with the glycidate epoxide synthesis, these processes are often described as "aza-Darzens reactions". This Microreview gives a summary of the area, with a focus on contemporary developments. ((C) Wiley-VCH Verlag GmbH & Co. KGaA, 69451 Weinheim, Germany, 2009)

Structure determination from powder diffraction data

Advances made over the past decade in structure determination from powder diffraction data are reviewed with particular emphasis on algorithmic developments and the successes and limitations of the technique. While global optimization methods have been successful in the solution of molecular crystal structures, new methods are required to make the solution of inorganic crystal structures more routine. The use of complementary techniques such as NMR to assist structure solution is discussed and the potential for the combined use of X-ray and neutron diffraction data for structure verification is explored. Structures that have proved difficult to solve from powder diffraction data are reviewed and the limitations of structure determination from powder diffraction data are discussed. Furthermore, the prospects of solving small protein crystal structures over the next decade are assessed.

Challenges and opportunities for robot-mediated neurorehabilitation

Robot-mediated neurorehabilitation is a rapidly advancing field that seeks to use advances in robotics, virtual realities, and haptic interfaces, coupled with theories in neuroscience and rehabilitation to define new methods for treating neurological injuries such as stroke, spinal cord injury, and traumatic brain injury. The field is nascent and much work is needed to identify efficient hardware, software, and control system designs alongside the most effective methods for delivering treatment in home and hospital settings. This paper identifies the need for robots in neurorehabilitation and identifies important goals that will allow this field to advance.

Remote sensing of cloud properties using ground-based measurements of zenith radiance

We have conducted the first extensive field test of two new methods to retrieve optical properties for overhead clouds that range from patchy to overcast. The methods use measurements of zenith radiance at 673 and 870 nm wavelengths and require the presence of green vegetation in the surrounding area. The test was conducted at the Atmospheric Radiation Measurement Program Oklahoma site during September–November 2004. These methods work because at 673 nm (red) and 870 nm (near infrared (NIR)), clouds have nearly identical optical properties, while vegetated surfaces reflect quite differently. The first method, dubbed REDvsNIR, retrieves not only cloud optical depth τ but also radiative cloud fraction. Because of the 1-s time resolution of our radiance measurements, we are able for the first time to capture changes in cloud optical properties at the natural timescale of cloud evolution. We compared values of τ retrieved by REDvsNIR to those retrieved from downward shortwave fluxes and from microwave brightness temperatures. The flux method generally underestimates τ relative to the REDvsNIR method. Even for overcast but inhomogeneous clouds, differences between REDvsNIR and the flux method can be as large as 50%. In addition, REDvsNIR agreed to better than 15% with the microwave method for both overcast and broken clouds. The second method, dubbed COUPLED, retrieves τ by combining zenith radiances with fluxes. While extra information from fluxes was expected to improve retrievals, this is not always the case. In general, however, the COUPLED and REDvsNIR methods retrieve τ to within 15% of each other.