Posterior-based proposals for speeding up Markov chain Monte Carlo.

Markov chain Monte Carlo (MCMC) is widely used for Bayesian inference in models of complex systems. Performance, however, is often unsatisfactory in models with many latent variables due to so-called poor mixing, necessitating the development of application-specific implementations. This paper introduces 'posterior-based proposals' (PBPs), a new type of MCMC update applicable to a huge class of statistical models (whose conditional dependence structures are represented by directed acyclic graphs). PBPs generate large joint updates in parameter and latent variable space, while retaining good acceptance rates (typically 33%). Evaluation against other approaches (from standard Gibbs/random walk updates to state-of-the-art Hamiltonian and particle MCMC methods) was carried out for widely varying model types: an individual-based model for disease diagnostic test data, a financial stochastic volatility model, a mixed model used in statistical genetics and a population model used in ecology. While different methods worked better or worse in different scenarios, PBPs were found to be either near to the fastest or significantly faster than the next best approach (by up to a factor of 10). PBPs, therefore, represent an additional general purpose technique that can be usefully applied in a wide variety of contexts.

Bayesian model evidence as a practical alternative to deviance information criterion.

While model evidence is considered by Bayesian statisticians as a gold standard for model selection (the ratio in model evidence between two models giving the Bayes factor), its calculation is often viewed as too computationally demanding for many applications. By contrast, the widely used deviance information criterion (DIC), a different measure that balances model accuracy against complexity, is commonly considered a much faster alternative. However, recent advances in computational tools for efficient multi-temperature Markov chain Monte Carlo algorithms, such as steppingstone sampling (SS) and thermodynamic integration schemes, enable efficient calculation of the Bayesian model evidence. This paper compares both the capability (i.e. ability to select the true model) and speed (i.e. CPU time to achieve a given accuracy) of DIC with model evidence calculated using SS. Three important model classes are considered: linear regression models, mixed models and compartmental models widely used in epidemiology. While DIC was found to correctly identify the true model when applied to linear regression models, it led to incorrect model choice in the other two cases. On the other hand, model evidence led to correct model choice in all cases considered. Importantly, and perhaps surprisingly, DIC and model evidence were found to run at similar computational speeds, a result reinforced by analytically derived expressions.

An analysis of effects of heterozygosity in dairy cattle for bovine tuberculosis resistance.

Genetic selection of cattle more resistant to bovine tuberculosis (bTB) may offer a complementary control strategy. Hypothesising underlying non-additive genetic variation, we present an approach using genome-wide high density markers to identify genomic loci with dominance effects on bTB resistance and to test previously published regions with heterozygote advantage in bTB. Our data comprised 1151 Holstein-Friesian cows from Northern Ireland, confirmed bTB cases and controls, genotyped with the 700K Illumina BeadChip. Genome-wide markers were tested for associations between heterozygosity and bTB status using marker-based relationships. Results were tested for robustness against genetic structure, and the genotypic frequencies of a significant locus were tested for departures from Hardy-Weinberg equilibrium. Genomic regions identified in our study and in previous publications were tested for dominance effects. Genotypic effects were estimated through ASReml mixed models. A SNP (rs43032684) on chromosome 6 was significant at the chromosome-wide level, explaining 1.7% of the phenotypic variance. In the controls, there were fewer heterozygotes for rs43032684 (P < 0.01) with the genotypic values suggesting that heterozygosity confers a heterozygote disadvantage. The region surrounding rs43032684 had a significant dominance effect (P < 0.01). SNP rs43032684 resides within a pseudogene with a parental gene involved in macrophage response to infection and within a copy-number-variation region previously associated with nematode resistance. No dominance effect was found for the region on chromosome 11, as indicated by a previous candidate region bTB study. These findings require further validation with large-scale data.

Using model-based proposals for fast parameter inference on discrete state space, continuous-time Markov processes.

Bayesian statistics provides a framework for the integration of dynamic models with incomplete data to enable inference of model parameters and unobserved aspects of the system under study. An important class of dynamic models is discrete state space, continuous-time Markov processes (DCTMPs). Simulated via the Doob-Gillespie algorithm, these have been used to model systems ranging from chemistry to ecology to epidemiology. A new type of proposal, termed 'model-based proposal' (MBP), is developed for the efficient implementation of Bayesian inference in DCTMPs using Markov chain Monte Carlo (MCMC). This new method, which in principle can be applied to any DCTMP, is compared (using simple epidemiological SIS and SIR models as easy to follow exemplars) to a standard MCMC approach and a recently proposed particle MCMC (PMCMC) technique. When measurements are made on a single-state variable (e.g. the number of infected individuals in a population during an epidemic), model-based proposal MCMC (MBP-MCMC) is marginally faster than PMCMC (by a factor of 2-8 for the tests performed), and significantly faster than the standard MCMC scheme (by a factor of 400 at least). However, when model complexity increases and measurements are made on more than one state variable (e.g. simultaneously on the number of infected individuals in spatially separated subpopulations), MBP-MCMC is significantly faster than PMCMC (more than 100-fold for just four subpopulations) and this difference becomes increasingly large.

Hydrodynamics of linked sphere model swimmers.

We describe in detail the hydrodynamics of a simple model of linked sphere swimmers. We calculate the asymptotic form of both the time averaged flow field generated by a single swimmer and the interactions between swimmers in a dilute suspension, showing how each depends on the parameters describing the swimmer and its swimming stroke. We emphasize the importance of time reversal symmetry in determining the far field flow around a swimmer and show that the interactions between swimmers are highly dependent on the relative phase of their swimming strokes.

Contact line dynamics in binary lattice Boltzmann simulations.

We show that, when a single relaxation time lattice Boltzmann algorithm is used to solve the hydrodynamic equations of a binary fluid for which the two components have different viscosities, strong spurious velocities in the steady state lead to incorrect results for the equilibrium contact angle. We identify the origins of these spurious currents and demonstrate how the results can be greatly improved by using a lattice Boltzmann method based on a multiple-relaxation-time algorithm. By considering capillary filling we describe the dependence of the advancing contact angle on the interface velocity.

Scattering of low-Reynolds-number swimmers.

We describe the consequences of time-reversal invariance of the Stokes equations for the hydrodynamic scattering of two low-Reynolds-number swimmers. For swimmers that are related to each other by a time-reversal transformation, this leads to the striking result that the angle between the two swimmers is preserved by the scattering. The result is illustrated for the particular case of a linked-sphere model swimmer. For more general pairs of swimmers, not related to each other by time reversal, we find that hydrodynamic scattering can alter the angle between their trajectories by several tens of degrees. For two identical contractile swimmers, this can lead to the formation of a bound state.

Lattice Boltzmann study of convective drop motion driven by nonlinear chemical kinetics.

We model a reaction-diffusion-convection system which comprises a liquid drop containing solutes that undergo an Oregonator reaction producing chemical waves. The reactants are taken to have surfactant properties so that the variation in their concentrations induces Marangoni flows at the drop interface which lead to a displacement of the drop. We discuss the mechanism by which the chemical-mechanical coupling leads to drop motion and the way in which the net displacement of the drop depends on the strength of the surfactant action. The equations of motion are solved using a lattice Boltzmann approach.

Capillary filling in patterned channels.

We show how the capillary filling of microchannels is affected by posts or ridges on the sides of the channels. Ridges perpendicular to the flow direction introduce contact line pinning, which slows, or sometimes prevents, filling, whereas ridges parallel to the flow provide extra surface that may enhance filling. Patterning the microchannel surface with square posts has little effect on the ability of a channel to fill for equilibrium contact angle theta_{e} less than approximately 30 degrees . For theta_{e} greater than approximately 60 degrees , however, even a small number of posts can pin the advancing liquid front.

Eliminating spurious velocities in the free-energy lattice Boltzmann method.

Spurious velocities are unphysical currents that appear close to curved interfaces in diffuse interface methods. We analyze the causes of these spurious velocities in the free-energy lattice Boltzmann algorithm. By making a suitable choice of the equilibrium distribution, and by finding the best way to numerically calculate derivatives, we show that these velocities may be decreased by an order of magnitude compared to previous models. Furthermore, we propose a momentum conserving forcing method that reduces spurious velocities by another factor of approximately 5. In three dimensions we find that 19 velocity vectors is the minimum number necessary.

Interface width and bulk stability: requirements for the simulation of deeply quenched liquid-gas systems.

Simulations of liquid-gas systems with interface terms evaluated by central difference discretizations are observed to fail to give accurate results for two reasons: the interface can get "stuck" on the lattice or a density overshoot develops around the interface. In the first case, the bulk densities can take a range of values, dependent on the initial conditions. In the second case, inaccurate bulk densities are found. We derived the minimum interface width required for the accurate simulation of liquid-gas systems with a diffuse interface. This criterion is demonstrated for lattice Boltzmann simulations of a van der Waals gas. Combining this criterion with predictions for the bulk stability defines the parameter range for stable and accurate simulation results even for high density ratios of over 1000.

Producing swimmers by coupling reaction-diffusion equations to a chemically responsive material.

We propose a mechanism for generating snakelike motion in a micrometer-scale, responsive, synthetic material, which thereby undergoes net movement in a fluid (i.e., swimming). By responsive material, we refer to a material that can expand or contract in response to a chemical concentration change. The concentrations of the chemical species are modeled by simple reaction-diffusion equations with suitably chosen source terms. Using linear stability analysis, we isolate the key properties of the material and reaction rate parameters.

Modeling microscopic swimmers at low Reynolds number.

The authors employ three numerical methods to explore the motion of low Reynolds number swimmers, modeling the hydrodynamic interactions by means of the Oseen tensor approximation, lattice Boltzmann simulations, and multiparticle collision dynamics. By applying the methods to a three bead linear swimmer, for which exact results are known, the authors are able to compare and assess the effectiveness of the different approaches. They then propose a new class of low Reynolds number swimmers, generalized three bead swimmers that can change both the length of their arms and the angle between them. Hence they suggest a design for a microstructure capable of moving in three dimensions. They discuss multiple bead, linear microstructures and show that they are highly efficient swimmers. They then turn to consider the swimming motion of elastic filaments. Using multiparticle collision dynamics the authors show that a driven filament behaves in a qualitatively similar way to the micron-scale swimming device recently demonstrated by Dreyfus et al. [Nature (London) 437, 862 (2005)].

Hydrodynamic interaction between two swimmers at low Reynolds number.

We investigate the hydrodynamic interactions between micro-organisms swimming at low Reynolds number. By considering simple model swimmers, and combining analytic and numerical approaches, we investigate the time-averaged flow field around a swimmer. At short distances the swimmer behaves like a pump. At large distances the velocity field depends on whether the swimming stroke is invariant under a combined time-reversal and parity transformation. We then consider two swimmers and find that the interaction between them consists of two parts: a passive term, independent of the motion of the second swimmer, and an active term resulting from the simultaneous swimming action of both swimmers. The swimmer-swimmer interaction is a complicated function of their relative displacement, orientation, and phase, leading to motion that can be attractive, repulsive, or oscillatory.

Pattern formation arising from condensation of a homogeneous gas into a binary, phase-separating liquid.

We examine the nucleated growth of a binary, immiscible liquid drop within a homogeneous gas. The system couples the growth of the liquid drop with the phase separation of the immiscible components and, thus, can potentially reveal novel pattern formation. To carry out this study, we first characterize the thermodynamic properties of the system in terms of an appropriate Ginzburg-Landau free energy density. By minimizing this free energy, we construct the equilibrium phase diagram for the system. We then use a lattice Boltzmann algorithm to solve the hydrodynamic equations describing the dynamical evolution of the fluid. We observe intriguing tentaclelike structures within the nucleation and growth regime and explore how the formation of these structures depends on the thermodynamic and transport properties of the system. We give scaling laws describing domain growth in both the diffusion- and flow-limited regimes. The results highlight the novel physics that can emerge when there is interplay between the ordering of a density and a concentration field.

Kinetics of the polymer collapse transition: the role of hydrodynamics.

We investigate numerically the dynamical behavior of a polymer chain collapsing in a dilute solution. The rate of collapse is measured with and without the presence of hydrodynamic interactions. We find that hydrodynamic interactions accelerate polymer collapse. We present a scaling theory describing the physical process responsible for the collapse kinetics. Predicted collapse times in a hydrodynamic (tauH approximately N(4/3)) and a Brownian heat bath (tauB approximately N2) agree well with the numerical results (tauH approximately N(1.40+/-0.08) and tauB approximately N(1.89+/-0.09)) where N denotes chain length. The folding kinetics of Go models of proteins is also examined. We show that for these systems, where many free energy minima compete, hydrodynamics has little effect on the kinetics.

Convection-driven pattern formation in phase-separating binary fluids.

Using a thermal-lattice Boltzmann model, we examine the rich phase behavior that develops when partially miscible fluids evolve in the presence of a vertical temperature gradient, which encompasses the critical temperature T(c) of the mixture. In particular, a binary AB fluid is confined between two plates in a gravitational field. The upper plate is fixed below T(c) and hence, the nearby fluid phase separates into A-rich and B-rich domains. The lower plate is fixed above the temperature T(c), and the surrounding fluid is in the homogeneous phase. A coupling between convection (driven by the temperature gradient) and phase separation gives rise to unique pattern formation. A number of regimes are identified: regularly spaced stripes, convective steady-state columns, the periodic disturbance of these columns, and finally, chaotic dripping from the upper surface. These results highlight dynamical behavior in partially miscible mixtures.

Kinetic theory derivation of the transport coefficients of stochastic rotation dynamics.

We use a kinetic theory approach to derive the continuum Navier-Stokes and heat conduction equations for stochastic rotation dynamics, a particle based algorithm for simulating a fluid. Hence we obtain expressions for the viscosity and thermal conductivity in two and three dimensions. The predictions are tested numerically and good agreement is found.

Pattern formation in binary fluids confined between rough, chemically heterogeneous surfaces.

Using a mesoscale model for hydrodynamics, we simulate driven flow of AB binary fluids past surfaces that contain well-defined roughness or asperities. The geometry and wetting properties of the asperities are found to have a dramatic effect on the flow patterns. We isolate conditions where the A fluid forms vertical bands that bridge the asperities and an imposed shear (or pressure gradient) drives the system to form monodisperse droplets of A within the B fluid. The size of the droplets can be tailored by varying the morphology of the asperities. The surfaces needed to create this rich dynamical behavior are used as the stamps in microcontact printing; thus, the parameter space can readily be accessed experimentally, and the predictions suggest an efficient method for forming emulsions with well-controlled morphologies.

Stripe formation in differentially forced binary systems.

We consider pattern formation in periodically forced binary systems. In particular, we focus on systems in which the two species are differentially forced, one being accelerated with respect to the other. Using a continuum model consisting of two isothermal ideal gases which interact via a frictional force we demonstrate analytically that stripes form spontaneously above a critical forcing amplitude. The wavelength of the stripes is found to be close to the wavelength of sound in the limit of small viscosity. The results are confirmed numerically. We suggest that the same mechanism may contribute to the formation of stripes in experiments on horizontally oscillated granular mixtures.