Statistical inference of regulatory networks for circadian regulation

Abstract

Abstract:We assess the accuracy of various state-of-the-art statistics and machine learning methods for reconstructing gene and protein regulatory networks in the context of circadian regulation. Our study draws on the increasing availability of gene expression and protein concentration time series for key circadian clock components in Arabidopsis thaliana. In addition, gene expression and protein concentration time series are simulated from a recently published regulatory network of the circadian clock in A. thaliana, in which protein and gene interactions are described by a Markov jump process based on Michaelis-Menten kinetics. We closely follow recent experimental protocols, including the entrainment of seedlings to different lightdark cycles and the knock-out of various key regulatory genes. Our study provides relative network reconstruction accuracy scores for a critical comparative performance evaluation, and sheds light on a series of highly relevant questions: it quantifies the influence of systematically missing values related to unknown protein concentrations and mRNA transcription rates, it investigates the dependence of the performance on the network topology and the degree of recurrency, it provides deeper insight into when and why non-linear methods fail to outperform linear ones, it offers improved guidelines on parameter settings in different inference procedures, and it suggests new hypotheses about the structure of the central circadian gene regulatory network in A. thaliana.

A sparse hierarchical Bayesian model for detecting relevant antigenic sites in virus evolution

“…For the mixed-effects elastic net we only show the results for α = 0.3 following Ruyssinck et al (2014), however the remaining results are available in Section 8 of the online supplementary materials. Alternative AUROC values based on using model selection and then ranking the variables based on the absolute values of the regression coefficients (Aderhold et al 2014), as well as other results, are also available in Section 8 of the online supplementary materials. Table 1 also measures the accuracy of predicting out of sample observations, y out , and the fixed effects coefficients, w in terms of Mean Squared Errors (MSEs).…”

Assessing the effect of unknown widespread perturbations in complex systems using the ν-gap

“…To assess the quality of our method, we carried out a comparison with the current state-of-the-art in network inference. We presented in the introduction the work of ( [1]): their analysis highlights that hierarchical Bayesian regression models (HBR) seem to perform best compared to most of the known methods. Hence we chose HBR for benchmarking.…”

“…Since the first application of Bayesian networks on gene expression data ( [8]), several network inference methods have been suggested, along with procedures to assess their accuracy ( [16], [5], [23]). A recent work has implemented a comparison of the most commonlyused network inference methods using both simulated and real data from the Arabidopsis circadian clock ( [1]). Methods were assessed according to a scoring system (AUROC) that evaluates the number of false positives and correct answers as a smooth function of the acceptance threshold.…”

“…Methods were assessed according to a scoring system (AUROC) that evaluates the number of false positives and correct answers as a smooth function of the acceptance threshold. The methods considered in the paper all infer causal networks, and include most of the well-known methodologies, such as Graphical Gaussian Models (GGM, [21], [7]), Sparse Regression (Lasso [22] and Elastic Net [24]), Time-varying Sparse Regression (Tesla [2]), Hierarchical Bayesian Regression models (HBR, [1]), Non-Homogeneous Hierarchical Bayesian models ( [11]), Automatic Relevance Determination in the context of Sparse Bayesian Regression (ARD-SBR, [20]), Bayesian Spline Autoregression (BSA, [17]), State Space Models (SSM, [3]), Gaussian processes (GP, [4]), Mutual information methods (ARACNE, [16]), Mixture Bayesian network models (MBN, [14]), and Gaussian Bayesian networks (BGe, [9]). Data were simulated using the computational model proposed by the Millar's group (Millar2010, [19]), in both deterministic and stochastic settings.…”

Approximate Bayesian inference in semi-mechanistic models

“…The first paradigm aims to apply generic models like sparse Lasso-type regression, Bayesian networks, or hierarchical Bayesian models. A recent overview and comparative evaluation was published by Aderhold et al (2014). The advantage of this approach is that the computational complexity of inference is comparatively low, and the application of these methods to problems of genuine interest is computationally feasible.…”