Nicola Soranzo

Comparing association network algorithms for reverse engineering of large-scale gene regulatory networks

MOTIVATIONnInferring a gene regulatory network exclusively from microarray expression profiles is a difficult but important task. The aim of this work is to compare the predictive power of some of the most popular algorithms in different conditions (like data taken at equilibrium or time courses) and on both synthetic and real microarray data. We are in particular interested in comparing similarity measures both of linear type (like correlations and partial correlations) and of non-linear type (mutual information and conditional mutual information), and in investigating the underdetermined case (less samples than genes).nnnRESULTSnIn our simulations we see that all network inference algorithms obtain better performances from data produced with structural perturbations, like gene knockouts at steady state, than with any dynamical perturbation. The predictive power of all algorithms is confirmed on a reverse engineering problem from Escherichia coli gene profiling data: the edges of the physical network of transcription factor-binding sites are significantly overrepresented among the highest weighting edges of the graph that we infer directly from the data without any structure supervision. Comparing synthetic and in vivo data on the same network graph allows us to give an indication of how much more complex a real transcriptional regulation program is with respect to an artificial model.nnnAVAILABILITYnSoftware is freely available at the URL http://people.sissa.it/~altafini/papers/SoBiAl07/.nnnSUPPLEMENTARY INFORMATIONnSupplementary data are available at Bioinformatics online.

From knockouts to networks: establishing direct cause-effect relationships through graph analysis.

Background Reverse-engineering gene networks from expression profiles is a difficult problem for which a multitude of techniques have been developed over the last decade. The yearly organized DREAM challenges allow for a fair evaluation and unbiased comparison of these methods. Results We propose an inference algorithm that combines confidence matrices, computed as the standard scores from single-gene knockout data, with the down-ranking of feed-forward edges. Substantial improvements on the predictions can be obtained after the execution of this second step. Conclusions Our algorithm was awarded the best overall performance at the DREAM4 In Silico 100-gene network sub-challenge, proving to be effective in inferring medium-size gene regulatory networks. This success demonstrates once again the decisive importance of gene expression data obtained after systematic gene perturbations and highlights the usefulness of graph analysis to increase the reliability of inference.

Discerning static and causal interactions in genome-wide reverse engineering problems

BACKGROUNDnIn the past years devising methods for discovering gene regulatory mechanisms at a genome-wide level has become a fundamental topic in the field of systems biology. The aim is to infer gene-gene interactions in an increasingly sophisticated and reliable way through the continuous improvement of reverse engineering algorithms exploiting microarray data.nnnMOTIVATIONnThis work is inspired by the several studies suggesting that coexpression is mostly related to static stable binding relationships, like belonging to the same protein complex, rather than other types of interactions more of a causal and transient nature (e.g. transcription factor-binding site interactions). The aim of this work is to verify if direct or conditional network inference algorithms (e.g. Pearson correlation for the former, partial Pearson correlation for the latter) are indeed useful in discerning static from causal dependencies in artificial and real gene networks (derived from Escherichia coli and Saccharomyces cerevisiae).nnnRESULTSnEven in the regime of weak inference power we have to work in, our analysis confirms the differences in the performances of the algorithms: direct methods are more robust in detecting stable interactions, conditional ones are better for causal interactions especially in presence of combinatorial transcriptional regulation.nnnSUPPLEMENTARY INFORMATIONnSupplementary data are available at Bioinformatics online.

ERNEST: a toolbox for chemical reaction network theory

SUMMARYnERNEST Reaction Network Equilibria Study Toolbox is a MATLAB package which, by checking various different criteria on the structure of a chemical reaction network, can exclude the multistationarity of the corresponding reaction system. The results obtained are independent of the rate constants of the reactions, and can be used for model discrimination.nnnAVAILABILITY AND IMPLEMENTATIONnThe software, implemented in MATLAB, is available under the GNU GPL free software license from http://people.sissa.it/ approximately altafini/papers/SoAl09/. It requires the MATLAB Optimization Toolbox.nnnCONTACTnaltafini@sissa.it.

Origin of Co-Expression Patterns in E.coli and S.cerevisiae Emerging from Reverse Engineering Algorithms

Background The concept of reverse engineering a gene network, i.e., of inferring a genome-wide graph of putative gene-gene interactions from compendia of high throughput microarray data has been extensively used in the last few years to deduce/integrate/validate various types of “physical” networks of interactions among genes or gene products. Results This paper gives a comprehensive overview of which of these networks emerge significantly when reverse engineering large collections of gene expression data for two model organisms, E.coli and S.cerevisiae, without any prior information. For the first organism the pattern of co-expression is shown to reflect in fine detail both the operonal structure of the DNA and the regulatory effects exerted by the gene products when co-participating in a protein complex. For the second organism we find that direct transcriptional control (e.g., transcription factor–binding site interactions) has little statistical significance in comparison to the other regulatory mechanisms (such as co-sharing a protein complex, co-localization on a metabolic pathway or compartment), which are however resolved at a lower level of detail than in E.coli. Conclusion The gene co-expression patterns deduced from compendia of profiling experiments tend to unveil functional categories that are mainly associated to stable bindings rather than transient interactions. The inference power of this systematic analysis is substantially reduced when passing from E.coli to S.cerevisiae. This extensive analysis provides a way to describe the different complexity between the two organisms and discusses the critical limitations affecting this type of methodologies.

mRNA stability and the unfolding of gene expression in the long-period yeast metabolic cycle

BackgroundIn yeast, genome-wide periodic patterns associated with energy-metabolic oscillations have been shown recently for both short (approx. 40 min) and long (approx. 300 min) periods.ResultsThe dynamical regulation due to mRNA stability is found to be an important aspect of the genome-wide coordination of the long-period yeast metabolic cycle. It is shown that for periodic genes, arranged in classes according either to expression profile or to function, the pulses of mRNA abundance have phase and width which are directly proportional to the corresponding turnover rates.ConclusionThe cascade of events occurring during the yeast metabolic cycle (and their correlation with mRNA turnover) reflects to a large extent the gene expression program observable in other dynamical contexts such as the response to stresses/stimuli.

Decompositions of large-scale biological systems based on dynamical properties

MOTIVATIONnGiven a large-scale biological network represented as an influence graph, in this article we investigate possible decompositions of the network aimed at highlighting specific dynamical properties.nnnRESULTSnThe first decomposition we study consists in finding a maximal directed acyclic subgraph of the network, which dynamically corresponds to searching for a maximal open-loop subsystem of the given system. Another dynamical property investigated is strong monotonicity. We propose two methods to deal with this property, both aimed at decomposing the system into strongly monotone subsystems, but with different structural characteristics: one method tends to produce a single large strongly monotone component, while the other typically generates a set of smaller disjoint strongly monotone subsystems.nnnAVAILABILITYnOriginal heuristics for the methods investigated are described in the article.nnnCONTACTnaltafini@sissa.it

LINEAR AND NONLINEAR METHODS FOR GENE REGULATORY NETWORK INFERENCE

Abstract In this work we compare the predictive power of some of the most popular algorithms used for gene network inference, seen as an unsupervised graph learning problem. The data, generated by an artificial model of a gene regulatory network, are taken in different conditions, like at equilibrium or during a time course, and different numbers of samples are considered in the reconstruction. For these data, we see that the performances of the algorithms are neatly superior for steady state data than for time series. Furthermore, we obtain that linear measures are better suited to capture linear behavior (like steady state conditions), while nonlinear measures are more effective for intrinsically nonlinear data (like the time course of our artificial network).

Simulation of the Benchmark Datasets

In this chapter, the in silico systems genetics dataset, used as a benchmark in the rest of the book, is described in detail, in particular regarding its simulation by SysGenSIM. Morever, the algorithms underlying the generation of the gene expression data and the genotype values are fully illustrated.

Modeling the genome-wide transient response to stimuli in yeast: Adaptation through integral feedback

The gene expression response of yeast to various types of stresses/perturbations shows a common pattern for the vast majority of genes, characterized by a quick transient peak followed by a return to the basal level (adaptation). In order to model this transient and the consequent adaptation, we use the idea of integral feedback (the integral representing the relative concentration of gene products). The resulting linear system with input explain sufficiently well the different time constants observable in the transient response while, at the same time, being in agreement with the known experimental degradation rates measurements.