Network Inference from Perturbation Time Course Data

Conference proceedings : ... Annual International Conference of the IEEE Engineering in Medicine and Biology Society. IEEE Engineering in Medicine and Biology Society. Annual Conference, 2009

Current methods for reconstructing biological networks often learn either the topology of large networks or the kinetic parameters of smaller networks with a well-characterized topology. We have recently described a network reconstruction algorithm, the Inferelator 1.0, that given a set of genome-wide measurements as input, simultaneously learns both topology and kinetic-parameters. Specifically, it learns a system of ordinary differential equations (ODEs) that describe the rate of change in transcription of each gene or gene-cluster, as a function of environmental and transcription factors. In order to scale to large networks, in Inferelator 1.0 we have approximated the system of ODEs to be uncoupled, and have solved each ODE using a one-step finite difference approximation. Naturally, these approximations become crude as the simulated time-interval increases. Here we present, implement, and test a new Markov-Chain-Monte-Carlo (MCMC) dynamical modeling method, Inferelator 2.0, that...

2018 IEEE Conference on Decision and Control (CDC), 2018

High-throughput data acquisition in synthetic biology leads to an abundance of data that need to be processed and aggregated into useful biological models. Building dynamical models based on this wealth of data is of paramount importance to understand and optimize designs of synthetic biology constructs. However, building models manually for each data set is inconvenient and might become infeasible for highly complex synthetic systems. In this paper, we present stateof-the-art system identification techniques and combine them with chemical reaction network theory (CRNT) to generate dynamic models automatically. On the system identification side, Sparse Bayesian Learning offers methods to learn from data the sparsest set of dictionary functions necessary to capture the dynamics of the system into ODE models; on the CRNT side, building on such sparse ODE models, all possible network structures within a given parameter uncertainty region can be computed. Additionally, the system identification process can be complemented with constraints on the parameters to, for example, enforce stability or non-negativity-thus offering relevant physical constraints over the possible network structures. In this way, the wealth of data can be translated into biologically relevant network structures, which then steers the data acquisition, thereby providing a vital step for closed-loop system identification.

Proceedings of the National Academy of Sciences, 2003

High-dimensional data sets generated by high-throughput technologies, such as DNA microarray, are often the outputs of complex networked systems driven by hidden regulatory signals. Traditional statistical methods for computing low-dimensional or hidden representations of these data sets, such as principal component analysis and independent component analysis, ignore the underlying network structures and provide decompositions based purely on a priori statistical constraints on the computed component signals. The resulting decomposition thus provides a phenomenological model for the observed data and does not necessarily contain physically or biologically meaningful signals. Here, we develop a method, called network component analysis, for uncovering hidden regulatory signals from outputs of networked systems, when only a partial knowledge of the underlying network topology is available. The a priori network structure information is first tested for compliance with a set of identifiability criteria. For networks that satisfy the criteria, the signals from the regulatory nodes and their strengths of influence on each output node can be faithfully reconstructed. This method is first validated experimentally by using the absorbance spectra of a network of various hemoglobin species. The method is then applied to microarray data generated from yeast Saccharamyces cerevisiae and the activities of various transcription factors during cell cycle are reconstructed by using recently discovered connectivity information for the underlying transcriptional regulatory networks.

This paper addresses the problem of identifying the large-scale topology of gene regulation networks from features that can be derived from microarray data sets. Understanding large-scale structures of gene regulation is fundamentally important in biology. Three main classes of network models, exponential network, scale-free network, and small-world network, have been used to describe topological features of various naturally occurring systems. Recent analysis of network properties of known biological networks have shown that they display scale free features, but it is not clear yet whether the scale free features are generic to all biological networks due to the limited availability of information on pathways and connectivity. To overcome these limitations and to expand the knowledge of biological network topology, we propose a novel method for network topology inference from microarray data. The proposed method is more robust than current reverse engineering approaches because it does not require inferring individual connectivity. Preliminary results with simulated data are encouraging. The trained neural network is able to classify networks as random network or scale free network with 90% accuracy and the mean connectivity can be predicted with around 85% to 90% accuracy. Our work shows that the neural network based topology inference method can predict the class of network topology and mean connectivity of network without knowing the underlying connectivity of structures from measured time series data.

BMC Systems Biology, 2012

Background High-throughput measurement technologies produce data sets that have the potential to elucidate the biological impact of disease, drug treatment, and environmental agents on humans. The scientific community faces an ongoing challenge in the analysis of these rich data sources to more accurately characterize biological processes that have been perturbed at the mechanistic level. Here, a new approach is built on previous methodologies in which high-throughput data was interpreted using prior biological knowledge of cause and effect relationships. These relationships are structured into network models that describe specific biological processes, such as inflammatory signaling or cell cycle progression. This enables quantitative assessment of network perturbation in response to a given stimulus. Results Four complementary methods were devised to quantify treatment-induced activity changes in processes described by network models. In addition, companion statistics were develop...

Journal of Theoretical Biology, 2005

One of the fundamental problems of cell biology is the understanding of complex regulatory networks. Such networks are ubiquitous in cells and knowledge of their properties is essential for the understanding of cellular behavior. In earlier work (Kholodenko et al. (PNAS 99: 12841), it was shown how the structure of biological networks can be quantified from experimental measurements of steady-state concentrations of key intermediates as a result of perturbations using a simple algorithm called ''unravelling''. Here, we study the effect of experimental uncertainty on the accuracy of the inferred structure (i.e. whether interactions are excitatory or inhibitory) of the networks determined using the unravelling algorithm. We show that the accuracy of the network structure depends not only on the noise level but on the strength of the interactions within the network. In particular, both very small and very large values of the connection strengths lead to large uncertainty in the inferred network. We describe a powerful geometric tool for the intuitive understanding of the effect of experimental error on the qualitative accuracy of the inferred network. In addition, we show that the use of additional data beyond that needed to minimally constrain the network not only improves the accuracy of the inferred network, but also may allow the detection of situations in which the initial assumptions of unravelling with respect to the network and the perturbations have been violated. Our ideas are illustrated using the mitogenactivated protein kinase (MAPK) signaling network as an example. r

2007

Background We consider the problem of identifying the dynamic interactions in biochemical networks from noisy experimental data. Typically, approaches for solving this problem make use of an estimation algorithm such as the well-known linear Least-Squares (LS) estimation technique.

IET Systems Biology, 2015

In computational systems biology, the general aim is to derive regulatory models from multivariate readouts, thereby generating predictions for novel experiments. In the past, many such models have been formulated for different biological applications. The authors consider the scenario where a given model fails to predict a set of observations with acceptable accuracy and ask the question whether this is because of the model lacking important external regulations. Real-world examples for such entities range from microRNAs to metabolic fluxes. To improve the prediction, they propose an algorithm to systematically extend the network by an additional latent dynamic variable which has an exogenous effect on the considered network. This variable's time course and influence on the other species is estimated in a two-step procedure involving spline approximation, maximum-likelihood estimation and model selection. Simulation studies show that such a hidden influence can successfully be inferred. The method is also applied to a signalling pathway model where they analyse real data and obtain promising results. Furthermore, the technique can be employed to detect incomplete network structures.

Computer Physics Communications, 2015

The dynamics of biological networks are fundamental to a variety of processes in many areas of biology and medicine. Understanding of such networks on a systemic level is facilitated by mathematical models describing these networks. However, since mathematical models of signalling networks commonly aim to describe several highly connected biological quantities and many model parameters cannot be measured directly, quantitative dynamic models often present challenges with respect to model calibration. Here, we propose an iterative fitting routine to decompose the problem of fitting a system of coupled ordinary differential equations describing a signalling network into smaller subproblems. Parameters for each differential equation are estimated separately using a Differential Evolution algorithm while all other dynamic quantities in the model are treated as input to the system. The performance of this algorithm is evaluated on artificial networks with known structure and known model parameters and compared to a conventional optimisation procedure for the same problem. Our analysis indicates that the procedure results in a significantly higher quality of fit and more efficient reconstruction of the true parameters than the conventional algorithm.

npj Systems Biology and Applications

Mathematical modeling of the emergent dynamics of gene regulatory networks (GRN) faces a double challenge of (a) dependence of model dynamics on parameters, and (b) lack of reliable experimentally determined parameters. In this paper we compare two complementary approaches for describing GRN dynamics across unknown parameters: (1) parameter sampling and resulting ensemble statistics used by RACIPE (RAndom CIrcuit PErturbation), and (2) use of rigorous analysis of combinatorial approximation of the ODE models by DSGRN (Dynamic Signatures Generated by Regulatory Networks). We find a very good agreement between RACIPE simulation and DSGRN predictions for four different 2- and 3-node networks typically observed in cellular decision making. This observation is remarkable since the DSGRN approach assumes that the Hill coefficients of the models are very high while RACIPE assumes the values in the range 1-6. Thus DSGRN parameter domains, explicitly defined by inequalities between systems p...