Circumventing structural uncertainty: A Bayesian perspective on nonlinear forecasting for ecology

As a consequence of the complexity of ecosystems and context-dependence of species interactions, structural uncertainty is pervasive in ecological modeling. This is particularly problematic when ecological models are used to make conservation and management plans whose outcomes may depend strongly on model formulation. Nonlinear time series approaches allow us to circumvent this issue by using the observed dynamics of the system to guide policy development. However, these methods typically require long time series from stationary systems, which are rarely available in ecological settings. Here we present a Bayesian approach to nonlinear forecasting based on Gaussian processes that readily integrates information from several short time series and allows for nonstationary dynamics. We demonstrate the utility of our modeling methods on simulated from a wide range of ecological scenarios. We expect that these models will extend the range of ecological systems to which nonlinear forecasting methods can be usefully applied.     

Modeling Dynamic Systems with Efficient Ensembles of Process-Based Models

Ensembles are a well established machine learning paradigm, leading to accurate and robust models, predominantly applied to predictive modeling tasks. Ensemble models comprise a finite set of diverse predictive models whose combined output is expected to yield an improved predictive performance as compared to an individual model. In this paper, we propose a new method for learning ensembles of process-based models of dynamic systems. The process-based modeling paradigm employs domain-specific knowledge to automatically learn models of dynamic systems from time-series observational data. Previous work has shown that ensembles based on sampling observational data (i.e., bagging and boosting), significantly improve predictive performance of process-based models. However, this improvement comes at the cost of a substantial increase of the computational time needed for learning. To address this problem, the paper proposes a method that aims at efficiently learning ensembles of process-based models, while maintaining their accurate long-term predictive performance. This is achieved by constructing ensembles with sampling domain-specific knowledge instead of sampling data. We apply the proposed method to and evaluate its performance on a set of problems of automated predictive modeling in three lake ecosystems using a library of process-based knowledge for modeling population dynamics. The experimental results identify the optimal design decisions regarding the learning algorithm. The results also show that the proposed ensembles yield significantly more accurate predictions of population dynamics as compared to individual process-based models. Finally, while their predictive performance is comparable to the one of ensembles obtained with the state-of-the-art methods of bagging and boosting, they are substantially more efficient.     

A new method of identifying a systems based on the minimal quadratic discrepancy criterion for biophysics problems

A wide range of biophysical systems are described by nonlinear dynamic models mathematically presented as a set of ordinary differential equations in the Cauchy explicit form: [formula: see text] Fij(X1(t),..,XN(t),t), (i = 1,...,N, j = 1,...,M), where Fij (X1(t), ..., XN(t), t) is a set of basis functions satisfying the Lipschitz condition. We investigate the problem of evaluation of model constants aij (the system identification) using experimental data about the time dependence of the dynamic parameters of the system Xi(t). A new method of system identification for the class of similar nonlinear dynamic models is proposed. It is shown that the problem of identifying an initial nonlinear model can be reduced to the solution of a system of linear equations for the matrix of the dynamic model constants [aj]i. It is proposed to determine the set of dynamic model constants aij using the criterion of minimal quadratic discrepancy for the time dependence of the set of dynamic parameters Xi(t). An important special case of the nonlinear model, the quadratic model, is considered. Test problems of identification using this method are presented for two nonlinear systems: the Van der Pol type multiparametric nonlinear oscillator and the strange attractor of Ressler, a widely known example of dynamic systems showing the stochastic behavior.     

Variables Aggregation in Time Varying Discrete Systems

In this work we extend approximate aggregation methods in time discrete linear models to the case of time varying environments. Approximate aggregation consists in describing some features of the dynamics of a general system involving many coupled variables in terms of the dynamics of a reduced system with a few number of variables. We present a time discrete time varying model in which we distinguish two time scales. By using perturbation methods we transform the system to make the global variables appear and build up the aggregated system. The asymptotic relationships between the general and aggregated systems are explored in the cases of a cyclically varying environment and a changing environment in process of stabilization. We show that under quite general conditions the knowledge of the behavior of the aggregated system characterizes that of the general system. The general method is also applied to aggregate a multiregional time dependent Leslie model showing that the aggregated model has demographic rates depending on the equilibrium proportions of individuals in the different patches.     

Integrating external biological knowledge in the construction of regulatory networks from time-series expression data

ABSTRACT: BACKGROUND: Inference about regulatory networks from high-throughput genomics data is of great interest in systems biology. We present a Bayesian approach to infer gene regulatory networks from time series expression data by integrating various types of biological knowledge. RESULTS: We formulate network construction as a series of variable selection problems and use linear regression to model the data. Our method summarizes additional data sources with an informative prior probability distribution over candidate regression models. We extend the Bayesian model averaging (BMA) variable selection method to select regulators in the regression framework. We summarize the external biological knowledge by an informative prior probability distribution over the candidate regression models. CONCLUSIONS: We demonstrate our method on simulated data and a set of time-series microarray experiments measuring the effect of a drug perturbation on gene expression levels, and show that it outperforms leading regression-based methods in the literature.     

A probabilistic methodology for integrating knowledge and experiments on biological networks.

Biological systems are traditionally studied by focusing on a specific subsystem, building an intuitive model for it, and refining the model using results from carefully designed experiments. Modern experimental techniques provide massive data on the global behavior of biological systems, and systematically using these large datasets for refining existing knowledge is a major challenge. Here we introduce an extended computational framework that combines formalization of existing qualitative models, probabilistic modeling, and integration of high-throughput experimental data. Using our methods, it is possible to interpret genomewide measurements in the context of prior knowledge on the system, to assign statistical meaning to the accuracy of such knowledge, and to learn refined models with improved fit to the experiments. Our model is represented as a probabilistic factor graph, and the framework accommodates partial measurements of diverse biological elements. We study the performance of several probabilistic inference algorithms and show that hidden model variables can be reliably inferred even in the presence of feedback loops and complex logic. We show how to refine prior knowledge on combinatorial regulatory relations using hypothesis testing and derive p-values for learned model features. We test our methodology and algorithms on a simulated model and on two real yeast models. In particular, we use our method to explore uncharacterized relations among regulators in the yeast response to hyper-osmotic shock and in the yeast lysine biosynthesis system. Our integrative approach to the analysis of biological regulation is demonstrated to synergistically combine qualitative and quantitative evidence into concrete biological predictions.     

结构方程模型及其在生态学中的应用

基于多变量统计方法同时研究自然系统内多个因子之间的相互关系, 是阐释复杂的自然系统的一个重要手段。相比传统的多变量统计法, 结构方程模型基于研究者的先验知识预先设定系统内因子间的依赖关系, 不仅能够判别各因子之间的关系强度(路径系数), 还能对整体模型进行拟合和判断, 从而能更全面地了解自然系统。由于结构方程模型只在近年才被应用到生态学的数据分析中, 因此该文试图对其作一简略介绍, 包括结构方程模型的定义和变量类型, 结合事例研究展现结构方程模型分析的一般步骤、在生态学中的应用以及相关软件的介绍等。望能为相关研究人员提供直观的认识, 加强结构方程模型在生态学数据分析中的应用。     

Towards building the silicon cell: a modular approach

Systems Biology aims to understand quantitatively how properties of biological systems can be understood as functions of the characteristics of, and interactions between their macromolecular components. Whereas, traditional biochemistry focused on isolation and characterization of cellular components, the challenge for Systems Biology lies in integration of this knowledge and the knowledge about molecular interactions. Computer models play an important role in this integration. We here discuss an approach with which we aim to link kinetic models on small parts of metabolism together, so as to form detailed kinetic models of larger chunks of metabolism, and ultimately of the entire living cell. Specifically, we will discuss techniques that can be used to model a sub-network in isolation of a larger network of which it is a part, while still maintaining the dynamics of the larger complete network. We will start by outlining the JWS online system, the silicon cell project, and the type of models we propose. JWS online is a model repository, which can be used for the storage, simulation and analysis of kinetic models. We advocate to integrate a top-down approach, where measurements on the complete system are used to derive fluxes in a detailed structural model, with a bottom-up approach, consisting of the integration of molecular mechanism-based detailed kinetic models into the structural model.     

Modeling and Estimation of Kinetic Parameters and Replicative Fitness of HIV-1 from Flow-Cytometry-Based Growth Competition Experiments

Growth competition assays have been developed to quantify the relative fitness of HIV-1 mutants. In this article, we develop mathematical models to describe viral/cellular dynamic interactions in the assay system from which the competitive fitness indices or parameters are defined. In our previous HIV-viral fitness experiments, the concentration of uninfected target cells was assumed to be constant (Wu et al. 2006). But this may not be true in some experiments. In addition, dual infection may frequently occur in viral fitness experiments and may not be ignorable. Here, we relax these two assumptions and extend our earlier viral fitness model (Wu et al. 2006). The resulting models then become nonlinear ODE systems for which closed-form solutions are not achievable. In the new model, the viral relative fitness is a function of time since it depends on the target cell concentration. First, we studied the structure identifiability of the nonlinear ODE models. The identifiability analysis showed that all parameters in the proposed models are identifiable from the flow-cytometry-based experimental data that we collected. We then employed a global optimization approach (the differential evolution algorithm) to directly estimate the kinetic parameters as well as the relative fitness index in the nonlinear ODE models using nonlinear least square regression based on the experimental data. Practical identifiability was investigated via Monte Carlo simulations.     

Image-based plant phenotyping with incremental learning and active contours

Plant phenotyping investigates how a plant's genome, interacting with the environment, affects the observable traits of a plant (phenome). It is becoming increasingly important in our quest towards efficient and sustainable agriculture. While sequencing the genome is becoming increasingly efficient, acquiring phenotype information has remained largely of low throughput. Current solutions for automated image-based plant phenotyping, rely either on semi-automated or manual analysis of the imaging data, or on expensive and proprietary software which accompanies costly hardware infrastructure. While some attempts have been made to create software applications that enable the analysis of such images in an automated fashion, most solutions are tailored to particular acquisition scenarios and restrictions on experimental design. In this paper we propose and test, a method for the segmentation and the automated analysis of time-lapse plant images from phenotyping experiments in a general laboratory setting, that can adapt to scene variability. The method involves minimal user interaction, necessary to establish the statistical experiments that may follow. At every time instance (i.e., a digital photograph), it segments the plants in images that contain many specimens of the same species. For accurate plant segmentation we propose a vector valued level set formulation that incorporates features of color intensity, local texture, and prior knowledge. Prior knowledge is incorporated using a plant appearance model implemented with Gaussian mixture models, which utilizes incrementally information from previously segmented instances. The proposed approach is tested on Arabidopsis plant images acquired with a static camera capturing many subjects at the same time. Our validation with ground truth segmentations and comparisons with state-of-the-art methods in the literature shows that the proposed method is able to handle images with complicated and changing background in an automated fashion. An accuracy of 96.7% (dice similarity coefficient) was observed, which was higher than other methods used for comparison. While here it was tested on a single plant species, the fact that we do not employ shape driven models and we do not rely on fully supervised classification (trained on a large dataset) increases the ease of deployment of the proposed solution for the study of different plant species in a variety of laboratory settings. Our solution will be accompanied by an easy to use graphical user interface and, to facilitate adoption, we will make the software available to the scientific community.     

Evolving cell models for systems and synthetic biology

This paper proposes a new methodology for the automated design of cell models for systems and synthetic biology. Our modelling framework is based on P systems, a discrete, stochastic and modular formal modelling language. The automated design of biological models comprising the optimization of the model structure and its stochastic kinetic constants is performed using an evolutionary algorithm. The evolutionary algorithm evolves model structures by combining different modules taken from a predefined module library and then it fine-tunes the associated stochastic kinetic constants. We investigate four alternative objective functions for the fitness calculation within the evolutionary algorithm: (1) equally weighted sum method, (2) normalization method, (3) randomly weighted sum method, and (4) equally weighted product method. The effectiveness of the methodology is tested on four case studies of increasing complexity including negative and positive autoregulation as well as two gene networks implementing a pulse generator and a bandwidth detector. We provide a systematic analysis of the evolutionary algorithm’s results as well as of the resulting evolved cell models.     

Robust optimization of total joint replacements incorporating environmental variables.

Direct search techniques for the optimal design of biomechanical devices are computationally intensive requiring many iterations before converging to a global solution. This, along with the incorporation of environmental variables such as multiple loading conditions and bone properties, makes direct search techniques infeasible. In this study, we introduced new methods that are based on the statistical design and analysis of computer experiments to account efficiently for environmental variables. Using data collected at a relatively small set of training sites, the method employs a computationally inexpensive predictor of the structural response that is statistically motivated. By using this predictor in place of the simulator (e.g., finite element model), a sufficient number of iterations can be performed to facilitate the optimization of the complex system. The applicability of these methods was demonstrated through the design of a femoral component for total hip arthroplasty incorporating variations in joint force orientation and cancellous bone properties. Beams on elastic foundation (BOEF) finite element models were developed to simulate the structural response. These simple models were chosen for their short computation time. This allowed us to represent the actual structural response surface by an exhaustive enumeration of the design and environmental variable space, and provided a means by which to validate the statistical predictor. We were able to predict the structural response and the optimal design accurately using only 16 runs of the computer code. The general trends predicted by the BOEF models were in agreement with previous three-dimensional finite element computer simulations, and experimental and clinical results, which demonstrated that the important features of intramedullary fixation systems were captured. These results indicate that the statistically based optimization methods are appropriate for optimization studies using computationally demanding models.     

The relationship of intercompartmental excitation transfer rate constants to those of an underlying physical model

In studies on photosynthetic systems it is common practice to interpret the results of time-resolved fluorescence experiments on the basis of compartmental, or target, models. Each compartment represents a group of molecules with similar fluorescence characteristics. In cases of practical interest, the members of each compartment are spatially contiguous and make up part of an overall energy-transferring system. Since a rate constant describing the overall transfer between compartments is not that of any pair of molecules in the system, this question naturally rises: what do we learn about the microscopic structure from these data? In this note we introduce ‘compartment melting’, a smooth mathematical connection between the compartmental and microscopic levels. We then show, on the basis of model calculations on finite lattices in one, two, and three dimensions, that average microscopic rates at the interfaces between compartments may be estimated from observed intercompartmental rates. The estimate involves a modest number of structural assumptions about the system. As examples of the method, which is applicable mainly to systems containing homogeneous pigment pools, some recent chlorophyll-protein antenna studies are analyzed.     

Modeling large RNAs and ribonucleoprotein particles using molecular mechanics techniques.

There is a growing body of low-resolution structural data that can be utilized to devise structural models for large RNAs and ribonucleoproteins. These models are routinely built manually. We introduce an automated refinement protocol to utilize such data for building low-resolution three-dimensional models using the tools of molecular mechanics. In addition to specifying the positions of each nucleotide, the protocol provides quantitative estimates of the uncertainties in those positions, i.e., the resolution of the model. In typical applications, the resolution of the models is about 10-20 A. Our method uses reduced representations and allows us to refine three-dimensional structures of systems as big as the 16S and 23S ribosomal RNAs, which are about one to two orders of magnitude larger than nucleic acids that can be examined by traditional all-atom modeling methods. Nonatomic resolution structural data--secondary structure, chemical cross-links, chemical and enzymatic footprinting patterns, protein positions, solvent accessibility, and so on--are combined with known motifs in RNA structure to predict low-resolution models of large RNAs. These structural constraints are imposed on the RNA chain using molecular mechanics-type potential functions with parameters based on the quality of experimental data. Surface potential functions are used to incorporate shape and positional data from electron microscopy image reconstruction experiments into our models. The structures are optimized using techniques of energy refinement to get RNA folding patterns. In addition to providing a consensus model, the method finds the range of models consistent with the data, which allows quantitative evaluation of the resolution of the model. The method also identifies conflicts in the experimental data. Although our protocol is aimed at much larger RNAs, we illustrate these techniques using the tRNA structure as an example and test-bed.     

Bridging the gap: linking molecular simulations and systemic descriptions of cellular compartments

Metabolic processes in biological cells are commonly either characterized at the level of individual enzymes and metabolites or at the network level. Often these two paradigms are considered as mutually exclusive because concepts from neither side are suited to describe the complete range of scales. Additionally, when modeling metabolic or regulatory cellular systems, often a large fraction of the required kinetic parameters are unknown. This even applies to such simple and extensively studied systems like the photosynthetic apparatus of purple bacteria. Using the chromatophore vesicles of Rhodobacter sphaeroides as a model system, we show that a consistent kinetic model emerges when fitting the dynamics of a molecular stochastic simulation to a set of time dependent experiments even though about two thirds of the kinetic parameters in this system are not known from experiment. Those kinetic parameters that were previously known all came out in the expected range. The simulation model was built from independent protein units composed of elementary reactions processing single metabolites. This pools-and-proteins approach naturally compiles the wealth of available molecular biological data into a systemic model and can easily be extended to describe other systems by adding new protein or nucleic acid types. The automated parameter optimization, performed with an evolutionary algorithm, reveals the sensitivity of the model to the value of each parameter and the relative importances of the experiments used. Such an analysis identifies the crucial system parameters and guides the setup of new experiments that would add most knowledge for a systemic understanding of cellular compartments. The successful combination of the molecular model and the systemic parametrization presented here on the example of the simple machinery for bacterial photosynthesis shows that it is actually possible to combine molecular and systemic modeling. This framework can now straightforwardly be applied to other currently less well characterized but biologically more relevant systems.     

The mechanism distinguishability problem in biochemical kinetics: the single-enzyme, single-substrate reaction as a case study

A theoretical analysis of the distinguishability problem of two rival models of the single enzyme-single substrate reaction, the Michaelis-Menten and Henri mechanisms, is presented. We also outline a general approach for analysing the structural indistinguishability between two mechanisms. The approach involves constructing, if possible, a smooth mapping between the two candidate models. Evans et al. [N.D. Evans, M.J. Chappell, M.J. Chapman, K.R. Godfrey, Structural indistinguishability between uncontrolled (autonomous) nonlinear analytic systems, Automatica 40 (2004) 1947-1953] have shown that if, in addition, either of the mechanisms satisfies a particular criterion then such a transformation always exists when the models are indistinguishable from their experimentally observable outputs. The approach is applied to the single enzyme-single substrate reaction mechanism. In principle, mechanisms can be distinguished using this analysis, but we show that our ability to distinguish mechanistic models depends both on the precise measurements made, and on our knowledge of the system prior to performing the kinetics experiments.     

Development and application of white-noise modeling techniques for studies of insect visual nervous system

The nonlinear system identification technique through white-noise stimulation is extended to multi-input, -output systems with consideration given to applications in the functional study of the nervous system. The applicability of the method is discussed in general and in particular for the motion detection neuronal system of the fly. Two series of experiments are performed; one with moving striped-pattern stimuli and the other with spot stimuli of fluctuating intensity. In both cases nonlinear dynamic models are derived which describe the system with considerable accuracy over the frequency range of 0.2–50 Hz and a dynamic amplitude range of about 40-1. These models are able to predict accurately all the discrete experiments so far performed on this system for which the models are applicable. The differences in dynamic characteristics between the corresponding system of the Musca and Phoenicia families of flies are minor except for a difference in latencies and if the difference in geometry of their faceted eyes is taken into account. The large field response of the motion detection unit is a linear weighted summation of all the smaller field highly nonlinear subsystems of which the large field is comprised.     

Building Structural Models of a Whole Mycoplasma Cell

Building structural models of entire cells has been a long-standing cross-discipline challenge for the research community, as it requires an unprecedented level of integration between multiple sources of biological data and enhanced methods for computational modeling and visualization. Here, we present the first 3D structural models of an entire Mycoplasma genitalium (MG) cell, built using the CellPACK suite of computational modeling tools. Our model recapitulates the data described in recent whole-cell system biology simulations and provides a structural representation for all MG proteins, DNA and RNA molecules, obtained by combining experimental and homology-modeled structures and lattice-based models of the genome. We establish a framework for gathering, curating and evaluating these structures, exposing current weaknesses of modeling methods and the boundaries of MG structural knowledge, and visualization methods to explore functional characteristics of the genome and proteome. We compare two approaches for data gathering, a manually-curated workflow and an automated workflow that uses homologous structures, both of which are appropriate for the analysis of mesoscale properties such as crowding and volume occupancy. Analysis of model quality provides estimates of the regularization that will be required when these models are used as starting points for atomic molecular dynamics simulations.