ARACNE: an algorithm for the reconstruction of gene regulatory networks in a mammalian cellular context.

Margolin AA, Nemenman I, Basso K, Wiggins C, Stolovitzky G, Dalla Favera R, Califano A. · 2006

Read ARACNE: an algorithm for the reconstruction of gene regulatory networks in a mammalian cellular context. by Margolin AA, Nemenman I, Basso K, Wiggins C, Stolovitzky G, Dalla Favera R, Califano A. (2006) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/europmc-pmc1810318.txt.

Introduction

Margolin, Califano and colleagues presented ARACNE in 2006, a method that rebuilds gene control networks in mammal cells. Earlier reverse-engineering tools worked only in simple organisms with small genomes. Co-expression links also mislead, since two genes can rise together through an indirect chain.

ARACNE uses information theory to strip out most of those indirect links. The authors prove it can rebuild a network exactly when feedback loops have negligible effect, and it held up well with many loops present. On realistic synthetic data it made very few errors and beat relevance networks and Bayesian networks. Applied to human B cell data, it recovered known control links in that immune cell type. The method was designed to scale to mammalian complexity.

Catalog record

CREATORMargolin AA, Nemenman I, Basso K, Wiggins C, Stolovitzky G, Dalla Favera R, Califano A.
YEAR2006
COLLECTIONPAPERS
FORMATTXT
FILE SIZE72 KB
LICENSECC BY 4.0
RIGHTSCC BY
SOURCEEurope PMC

Source: https://europepmc.org/articles/PMC1810318

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

ARACNE: An Algorithm for the Reconstruction of Gene Regulatory Networks in a Mammalian Cellular Context By Adam A Margolin, Ilya Nemenman, Katia Basso, Chris Wiggins, Gustavo Stolovitzky, Riccardo Dalla Favera, Andrea Califano Published in: BMC Bioinformatics DOI: 10.1186/1471-2105-7-S1-S7 ABSTRACT Background Elucidating gene regulatory networks is crucial for understanding normal cell physiology and complex pathologic phenotypes. Existing computational methods for the genome-wide "reverse engineering" of such networks have been successful only for lower eukaryotes with simple genomes. Here we present ARACNE, a novel algorithm, using microarray expression profiles, specifically designed to scale up to the complexity of regulatory networks in mammalian cells, yet general enough to address a wider range of network deconvolution problems. This method uses an information theoretic approach to eliminate the majority of indirect interactions inferred by co-expression methods. Results We prove that ARACNE reconstructs the network exactly (asymptotically) if the effect of loops in the network topology is negligible, and we show that the algorithm works well in practice, even in the presence of numerous loops and complex topologies. We assess ARACNE's ability to reconstruct transcriptional regulatory networks using both a realistic synthetic dataset and a microarray dataset from human B cells. On synthetic datasets ARACNE achieves very low error rates and outperforms established methods, such as Relevance Networks and Bayesian Networks. Application to the deconvolution of genetic networks in human B cells demonstrates ARACNE's ability to infer validated transcriptional targets of the cMYC proto-oncogene. We also study the effects of misestimation of mutual information on network reconstruction, and show that algorithms based on mutual information ranking are more resilient to estimation errors. Conclusion ARACNE shows promise in identifying direct transcriptional interactions in mammalian cellular networks, a problem that has challenged existing reverse engineering algorithms. This approach should enhance our ability to use microarray data to elucidate functional mechanisms that underlie cellular processes and to identify molecular targets of pharmacological compounds in mammalian cellular networks. BACKGROUND Cellular phenotypes are determined by the dynamical activity of large networks of co-regulated genes. Thus dissecting the mechanisms of phenotypic selection requires elucidating the functions of the individual genes in the context of the networks in which they operate. Because gene expression is regulated by proteins, which are themselves gene products, statistical associations between gene mRNA abundance levels, while not directly proportional to activated protein concentrations, should provide clues towards uncovering gene regulatory mechanisms. Consequently, the advent of high throughput microarray technologies to simultaneously measure mRNA abundance levels across an entire genome has spawned much research aimed at using these data to construct conceptual "gene network" models to concisely describe the regulatory influences that genes exert on each other. Genome-wide clustering of gene expression profiles [1] provides an important first step towards this goal by grouping together genes that exhibit similar transcriptional responses to various cellular conditions, and are therefore likely to be involved in similar cellular processes. However, the organization of genes into co-regulated clusters provides a very coarse representation of the cellular network. In particular, it cannot separate statistical interactions that are irreducible (i.e., direct) from those arising from cascades of transcriptional interactions that correlate the expression of many noninteracting genes. More generally, as appreciated in statistical physics, long range order (i.e., high correlation among non-directly interacting variables) can easily result from short range interactions [2]. Thus correlations, or any other local dependency measure, cannot be used as the only tool for the reconstruction of interaction networks without additional assumptions. Within the last few years a number of sophisticated approaches for the reverse engineering of cellular networks (also called deconvolution) from gene expression data have emerged (reviewed in [3]). Their goal is to produce a high-fidelity representation of the cellular network topology as a graph, where genes are represented as vertices and are connected by edges representing direct regulatory interactions. The criteria for defining an edge, as well as its biological interpretation, remain imprecise and vary between applications. For example, graphical modeling [4] defines edges as parent-child relationships between mRNA abundance levels that are most likely to explain the data, integrative methods [5] use independent experimental clues to define edges as those showing evidence of physical interactions, and other statistical/information theoretical methods [6] identify edges with the strongest statistical associations between mRNA abundance levels. All available approaches suffer to various degrees from problems such as overfitting, high computational complexity, reliance on non-realistic network models, or a critical dependency on supplementary data that is only available for simple organisms. These limitations have relegated the successful large-scale application of most methods to relatively simple organisms, such as the yeast Saccharomyces cerevisiae, and the genome-wide deconvolution of a mammalian network is yet to be reported. Here we introduce ARACNE (Algorithm for the Reconstruction of Accurate Cellular Networks), a novel information-theoretic algorithm for the reverse engineering of transcriptional networks from microarray data that overcomes some of these limitations. ARACNE defines an edge as an irreducible statistical dependency between gene expression profiles that cannot be explained as an artifact of other statistical dependencies in the network. We suggest that the presence of such irreducible statistical dependencies is likely to identify direct regulatory interactions mediated by a transcription factor binding to a target gene's promoter region, although other types of interactions may also be identified (see Discussion). In this study we focus on the former type of interaction for validation purposes. We demonstrate that ARACNE compares favorably with existing methods and achieves extremely low error rates in identifying transcriptional interactions in a synthetic dataset modeled using realistic Hill kinetics. In a biological context, we show that the algorithm infers bona-fide transcriptional targets in a mammalian gene network. We also study the effects of misestimation of mutual information (MI) on network reconstruction, and show that algorithms based on MI ranking are resilient to estimation errors. The algorithm is general enough to deal with a variety of other network reconstruction problems in biological, social, and engineering fields. Theoretical Background Several factors have impeded the reliable reconstruction of genome-wide mammalian networks. First, temporal gene expression data is difficult to obtain for higher eukaryotes, and cellular populations harvested from different individuals generally capture random steady states of the underlying biochemical dynamics. This precludes the use of methods that infer temporal associations and thus plausible causal relationships (reviewed in [7]). Only steady state statistical dependences can be studied, which are not obviously linked to the underlying physical dependency model. Compounding this constraint, there is no universally accepted definition of statistical dependencies in the multivariate setting [8,9]. In this work we adopt the definition of [9], which builds on ideas from the Markov networks literature [10]. Briefly, we write the joint probability distribution (JPD) of the stationary expressions of all genes, P({gi}), i = 1,..., N, as: P({gi})=1Zexp⁡[−∑iNφi(gi)−∑i,jNφij(gi,gj)−∑i,j,kNφijk(gi,gj,gk)−...]≡e−H({gi}) (1) MathType@MTEF@5@5@+=feaafiart1ev1aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBaebbnrfifHhDYfgasaacH8akY=wiFfYdH8Gipec8Eeeu0xXdbba9frFj0=OqFfea0dXdd9vqai=hGuQ8kuc9pgc9s8qqaq=dirpe0xb9q8qiLsFr0=vr0=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@8F9F@ where N is the number of genes, Z is the normalization factor, also called the partition function, φ... are potentials, and H({gi}) is the Hamiltonian that defines the system's statistics. Within such a model, we assert that a set of variables interacts if and only if (iff) the single potential that depends exclusively on these variables is nonzero. ARACNE aims precisely at identifying which of these potentials are nonzero, and eliminating the others even though their corresponding marginal JPDs may not

Read the complete text