Jean-Loup Faulon

Predicting protein--protein interactions using signature products

MOTIVATION Proteome-wide prediction of protein-protein interaction is a difficult and important problem in biology. Although there have been recent advances in both experimental and computational methods for predicting protein-protein interactions, we are only beginning to see a confluence of these techniques. In this paper, we describe a very general, high-throughput method for predicting protein-protein interactions. Our method combines a sequence-based description of proteins with experimental information that can be gathered from any type of protein-protein interaction screen. The method uses a novel description of interacting proteins by extending the signature descriptor, which has demonstrated success in predicting peptide/protein binding interactions for individual proteins. This descriptor is extended to protein pairs by taking signature products. The signature product is implemented within a support vector machine classifier as a kernel function. RESULTS We have applied our method to publicly available yeast, Helicobacter pylori, human and mouse datasets. We used the yeast and H.pylori datasets to verify the predictive ability of our method, achieving from 70 to 80% accuracy rates using 10-fold cross-validation. We used the human and mouse datasets to demonstrate that our method is capable of cross-species prediction. Finally, we reused the yeast dataset to explore the ability of our algorithm to predict domains. CONTACT smartin@sandia.gov

A retrosynthetic biology approach to metabolic pathway design for therapeutic production

BackgroundSynthetic biology is used to develop cell factories for production of chemicals by constructively importing heterologous pathways into industrial microorganisms. In this work we present a retrosynthetic approach to the production of therapeutics with the goal of developing an in situ drug delivery device in host cells. Retrosynthesis, a concept originally proposed for synthetic chemistry, iteratively applies reversed chemical transformations (reversed enzyme-catalyzed reactions in the metabolic space) starting from a target product to reach precursors that are endogenous to the chassis. So far, a wider adoption of retrosynthesis into the manufacturing pipeline has been hindered by the complexity of enumerating all feasible biosynthetic pathways for a given compound.ResultsIn our method, we efficiently address the complexity problem by coding substrates, products and reactions into molecular signatures. Metabolic maps are represented using hypergraphs and the complexity is controlled by varying the specificity of the molecular signature. Furthermore, our method enables candidate pathways to be ranked to determine which ones are best to engineer. The proposed ranking function can integrate data from different sources such as host compatibility for inserted genes, the estimation of steady-state fluxes from the genome-wide reconstruction of the organisms metabolism, or the estimation of metabolite toxicity from experimental assays. We use several machine-learning tools in order to estimate enzyme activity and reaction efficiency at each step of the identified pathways. Examples of production in bacteria and yeast for two antibiotics and for one antitumor agent, as well as for several essential metabolites are outlined.ConclusionsWe present here a unified framework that integrates diverse techniques involved in the design of heterologous biosynthetic pathways through a retrosynthetic approach in the reaction signature space. Our engineering methodology enables the flexible design of industrial microorganisms for the efficient on-demand production of chemical compounds with therapeutic applications.

Enumerating metabolic pathways for the production of heterologous target chemicals in chassis organisms

BackgroundWe consider the possibility of engineering metabolic pathways in a chassis organism in order to synthesize novel target compounds that are heterologous to the chassis. For this purpose, we model metabolic networks through hypergraphs where reactions are represented by hyperarcs. Each hyperarc represents an enzyme-catalyzed reaction that transforms set of substrates compounds into product compounds. We follow a retrosynthetic approach in order to search in the metabolic space (hypergraphs) for pathways (hyperpaths) linking the target compounds to a source set of compounds.ResultsTo select the best pathways to engineer, we have developed an objective function that computes the cost of inserting a heterologous pathway in a given chassis organism. In order to find minimum-cost pathways, we propose in this paper two methods based on steady state analysis and network topology that are to the best of our knowledge, the first to enumerate all possible heterologous pathways linking a target compounds to a source set of compounds. In the context of metabolic engineering, the source set is composed of all naturally produced chassis compounds (endogenuous chassis metabolites) and the target set can be any compound of the chemical space. We also provide an algorithm for identifying precursors which can be supplied to the growth media in order to increase the number of ways to synthesize specific target compounds.ConclusionsWe find the topological approach to be faster by several orders of magnitude than the steady state approach. Yet both methods are generally scalable in time with the number of pathways in the metabolic network. Therefore this work provides a powerful tool for pathway enumeration with direct application to biosynthetic pathway design.

Molecular signatures-based prediction of enzyme promiscuity

MOTIVATION Enzyme promiscuity, a property with practical applications in biotechnology and synthetic biology, has been related to the evolvability of enzymes. At the molecular level, several structural mechanisms have been linked to enzyme promiscuity in enzyme families. However, it is at present unclear to what extent these observations can be generalized. Here, we introduce for the first time a method for predicting catalytic and substrate promiscuity using a graph-based representation known as molecular signature. RESULTS Our method, which has an accuracy of 85% for the non-redundant KEGG database, is also a powerful analytical tool for characterizing structural determinants of protein promiscuity. Namely, we found that signatures with higher contribution to the prediction of promiscuity are uniformly distributed in the protein structure of promiscuous enzymes. In contrast, those signatures that act as promiscuity determinants are significantly depleted around non-promiscuous catalytic sites. In addition, we present the study of the enolase and aminotransferase superfamilies as illustrative examples of characterization of promiscuous enzymes within a superfamily and achievement of enzyme promiscuity by protein reverse engineering. Recognizing the role of enzyme promiscuity in the process of natural evolution of enzymatic function can provide useful hints in the design of directed evolution experiments. We have developed a method with potential applications in the guided discovery and enhancement of latent catalytic capabilities surviving in modern enzymes. AVAILABILITY http://www.issb.genopole.fr~faulon.

Retropath: Automated Pipeline for Embedded Metabolic Circuits

Metabolic circuits are a promising alternative to other conventional genetic circuits as modular parts implementing functionalities required for synthetic biology applications. To date, metabolic design has been mainly focused on production circuits. Emergent applications such as smart therapeutics, however, require circuits that enable sensing and regulation. Here, we present RetroPath, an automated pipeline for embedded metabolic circuits that explores the circuit design space from a given set of specifications and selects the best circuits to implement based on desired constraints. Synthetic biology circuits embedded in a chassis organism that are capable of controlling the production, processing, sensing, and the release of specific molecules were enumerated in the metabolic space through a standard procedure. In that way, design and implementation of applications such as therapeutic circuits that autonomously diagnose and treat disease, are enabled, and their optimization is streamlined.

OMG: Open Molecule Generator

Computer Assisted Structure Elucidation has been used for decades to discover the chemical structure of unknown compounds. In this work we introduce the first open source structure generator, Open Molecule Generator (OMG), which for a given elemental composition produces all non-isomorphic chemical structures that match that elemental composition. Furthermore, this structure generator can accept as additional input one or multiple non-overlapping prescribed substructures to drastically reduce the number of possible chemical structures. Being open source allows for customization and future extension of its functionality. OMG relies on a modified version of the Canonical Augmentation Path, which grows intermediate chemical structures by adding bonds and checks that at each step only unique molecules are produced. In order to benchmark the tool, we generated chemical structures for the elemental formulas and substructures of different metabolites and compared the results with a commercially available structure generator. The results obtained, i.e. the number of molecules generated, were identical for elemental compositions having only C, O and H. For elemental compositions containing C, O, H, N, P and S, OMG produces all the chemically valid molecules while the other generator produces more, yet chemically impossible, molecules. The chemical completeness of the OMG results comes at the expense of being slower than the commercial generator. In addition to being open source, OMG clearly showed the added value of constraining the solution space by using multiple prescribed substructures as input. We expect this structure generator to be useful in many fields, but to be especially of great importance for metabolomics, where identifying unknown metabolites is still a major bottleneck.

Stereo Signature Molecular Descriptor

We present an algorithm to compute molecular graph descriptors considering the stereochemistry of the molecular structure based on our previously introduced signature molecular descriptor. The algorithm can generate two types of descriptors, one which is compliant with the Cahn-Ingold-Prelog priority rules, including complex stereochemistry structures such as fullerenes, and a computationally efficient one based on our previous definition of a directed acyclic graph that is augmented to a chiral molecular graph. The performance of the algorithm in terms of speed as a canonicalizer as well as in modeling and predicting bioactivity is evaluated, showing an overall better performance than other molecular descriptors, which is particularly relevant in modeling stereoselective biochemical reactions. The complete source code of the stereo signature molecular descriptor is available for download under an open-source license at http://molsig.sourceforge.net.

Origins of Specificity and Promiscuity in Metabolic Networks

Background: How enzymes evolved to their present form is linked to how extant metabolic pathways emerged. Results: Chemical diversity of reactions parallels enzyme phylogenetic diversity across the tree of life. Conclusion: Enzyme promiscuity plays a prominent role in the evolution of metabolic networks. Significance: Learning about the mechanisms of enzyme evolution might assist us with the identification of primeval catalytic functions and minimal metabolism. How enzymes have evolved to their present form is linked to the question of how pathways emerged and evolved into extant metabolic networks. To investigate this mechanism, we have explored the chemical diversity present in a largely unbiased data set of catalytic reactions processed by modern enzymes across the tree of life. In order to get a quantitative estimate of enzyme chemical diversity, we measure enzyme multispecificity or promiscuity using the reaction molecular signatures. Our main finding is that reactions that are catalyzed by a highly specific enzyme are shared by poorly divergent species, suggesting a later emergence of this function during evolution. In contrast, reactions that are catalyzed by highly promiscuous enzymes are more likely to appear uniformly distributed across species in the tree of life. From a functional point of view, promiscuous enzymes are mainly involved in amino acid and lipid metabolisms, which might be associated with the earliest form of biochemical reactions. In this way, results presented in this paper might assist us with the identification of primeval promiscuous catalytic functions contributing to lifes minimal metabolism.

Compound toxicity screening and structure–activity relationship modeling in Escherichia coli†

Synthetic biology and metabolic engineering are used to develop new strategies for producing valuable compounds ranging from therapeutics to biofuels in engineered microorganisms. When developing methods for high-titer production cells, toxicity is an important element to consider. Indeed the production rate can be limited due to toxic intermediates or accumulation of byproducts of the heterologous biosynthetic pathway of interest. Conversely, highly toxic molecules are desired when designing antimicrobials. Compound toxicity in bacteria plays a major role in metabolic engineering as well as in the development of new antibacterial agents. Here, we screened a diversified chemical library of 166 compounds for toxicity in Escherichia coli. The dataset was built using a clustering algorithm maximizing the chemical diversity in the library. The resulting assay data was used to develop a toxicity predictor that we used to assess the toxicity of metabolites throughout the metabolome. This new tool for predicting toxicity can thus be used for fine-tuning heterologous expression and can be integrated in a computational-framework for metabolic pathway design. Many structure-activity relationship tools have been developed for toxicology studies in eukaryotes [Valerio (2009), Toxicol Appl Pharmacol, 241(3): 356-370], however, to the best of our knowledge we present here the first E. coli toxicity prediction web server based on QSAR models (EcoliTox server: http://www.issb.genopole.fr/∼faulon/EcoliTox.php).

Engineering antibiotic production and overcoming bacterial resistance.

Progress in DNA technology, analytical methods and computational tools is leading to new developments in synthetic biology and metabolic engineering, enabling new ways to produce molecules of industrial and therapeutic interest. Here, we review recent progress in both antibiotic production and strategies to counteract bacterial resistance to antibiotics. Advances in sequencing and cloning are increasingly enabling the characterization of antibiotic biosynthesis pathways, and new systematic methods for de novo biosynthetic pathway prediction are allowing the exploration of the metabolic chemical space beyond metabolic engineering. Moreover, we survey the computer‐assisted design of modular assembly lines in polyketide synthases and non‐ribosomal peptide synthases for the development of tailor‐made antibiotics. Nowadays, production of novel antibiotic can be tranferred into any chosen chassis by optimizing a host factory through specific strain modifications. These advances in metabolic engineering and synthetic biology are leading to novel strategies for engineering antimicrobial agents with desired specificities.