Programme INRIA "Equipes Associées"
/ INRIA "Associate Teams" Programme

EQUIPE ASSOCIEE / ASSOCIATE TEAM	AMIGO
sélection	2010

 

Equipe-Projet INRIA : BAMBOO	Organisme étranger partenaire / Partner Institution: INESC-ID IST Lisbon (KDBio Team)
Centre de recherche INRIA : Rhône-Alpes Grenoble Thème INRIA : BIO A	Pays / Country : Portugal

 

	Coordinateur français / French Coordinator	Coordinateur étranger / Partner Coordinator
Nom, prénom / First name, Given name	SAGOT, Marie-France	FREITAS, Ana Teresa
Grade, statut / Position	Research Director (DR2)	Assistant Professor (with tenure) / Senior Researcher
Organisme d'appartenance / Home Institution (précisez le département et/ou le laboratoire)	INRIA Rhône-Alpes Grenoble	Dept. of Electrical and Computer Engineering, Technical University of Lisbon (IST) / INESC-ID KDBio Team
Adresse postale / Postal address	UMR 5558 LBBE Université Claude Bernard, Lyon I - Bât Mendel 43, Bd du 11 novembre 1918 69622 Villeurbanne cedex France	INESC-ID Rua Alves Redol, 9 1000-029 Lisboa Portugal
URL / Website	http://pbil.univ-lyon1.fr/members/sagot/htdocs/index.html	http://kdbio.inesc-id.pt/~atf
Téléphone / Telephone	+33 (0)4 72 44 82 38	+351 213100394
Télécopie / Fax	+33 (0)4 72 43 13 88	+351 213145843
Courriel / Email	Marie-France.Sagot@inria.fr	atf@inesc-id.pt

Titre de la thématique de collaboration (en français et en anglais) / Title of the collaboration theme (in French and in English) : Computational Methods for the Inference and Analysis of Biological Networks Méthodes Informatiques pour l'Inférence et Analyse des Réseaux Biologiques

Descriptif (environ 10 lignes) / Description (approximately 10 lines) : Biological systems can be described in terms of an overlay of networks, each providing a partial view on the elementary constituents and interactions that define the complex structure and dynamics of such systems. However, the assembly of the diverse data currently available into biologically sound networks leading to realistic predictive models remains a daunting computational task, requiring the conception of adequate mathematical models and efficient data structures and algorithms for their representation, inference, and downstream analysis. In this project, we focus on two points: (i) the identification and target prediction of small RNAs (notably miRNAs), which have been recently found to be important, though often neglected, actors of regulation networks, and (ii) the exploration of interconnections between structure and dynamics of metabolic networks. We intend to foster the already-existing collaboration between the partners (specially in the first subject) as well as to create new synergy, by exploring the complementarity of their technical expertise (specially on the second theme).

Présentation détaillée de l'Équipe Associée
Detailed presentation of the Associate Team

1. Objectifs scientifiques de la proposition (1 à 2 pages)
/ Scientific goals of the proposal (1 to 2 pages)

This collaboration will be carried out on two main topics: (Part I) the development of algorithms and models for predicting small non-coding RNAs, and (Part II) the exploration of the interconnections between structure and dynamics of metabolic networks.

Description of Part I

The discovery of small RNAs changed the perspective on eukaryotic gene expression regulation. It is now becoming apparent that these regulators have a broad and important role in the regulation of a number of developmental processes in animals and plants. In this project we shall develop tools for the identification and characterisation of small non-coding RNAs, in particular MicroRNAs (miRNAs). Although many methods exist for identifying miRNAs or their targets, the relatively high level of false positives and false negatives still observed illustrates our continued ignorance of the subtle rules presiding miRNA biogenesis and target specificity (Mendes et al., Nucleic Acids Res 2009, 37(8):2419-33)^*. Developing methods that enable to find and to analyse miRNA genes and miRNA-target associations based on more accurate models is therefore both a computational and a bioinformatics challenge.

The problems of miRNA gene finding and target prediction fall within the more general objective of unraveling the network of miRNA-mediated gene regulation in an organism of interest. This network is thought to be made up of miRNA regulatory modules in which a subset of an organism's miRNA repertoire is jointly involved in the silencing of a subset of protein-coding genes, presumably with a common and identifiable biological function. The problem of identifying such regulatory modules has not been addressed directly, but it has rather been occasionally dealt with as an additional step after the use of miRNA gene finding and target prediction tools, applied independently. With this research we shall develop a framework that constitutes an evolving method, able to draw from heterogeneous sources of data and to offer an evidence-based tool to estimate the likelihood that a given set of miRNA candidates and miRNA-target possible associations constitute a miRNA-mediated regulatory module.

In this research, we shall deal with a set of insect genomes and a set of human genomes. Despite the fact that the biological problems that we shall consider are very different, the methodological approach will be the same for both data sets: the set of miRNA precursor candidates and miRNAs candidates will be obtained both from genome-wide search and from high-throughput sequencing of small RNAs, using the 454 GS FLX platform.

Regarding the insect data set, a genome-wide search will be performed considering the genomes of the mosquitoes Anopheles gambiae and Anopheles darlingi, the primary malaria vectors in Africa and in Latin America, respectively, as well as that of the fruit fly Drosophila melanogaster, the main insect model organism. This study seeks to characterise the role of insect miRNAs in the response to parasite infection. It is important to note that the genome of Anopheles darlingi was recently sequenced at the LNCC (National Laboratory of Scientific Computing) research institute in Petrópolis, Brazil and will be analysed for the first time in the context of a collaboration research project both partners have with the LNCC.

As for the human data, the expression profile of the miRNAs will be determined by high-throughput sequencing, using the 454 GS FLX platform, with subsequent confirmation by real-time quantitative PCR. Prediction of the miRNAs and corresponding potential targets will be done by the algorithms developed in the scope of this project. We plan to identify miRNAs that have as specific targets molecules that are related to cell migration, whose computationally predicted messenger RNA (mRNA) targets suggest a control of the expression of such transcripts. Determining which miRNAs are differentially expressed in T cell lymphomas when compared to normal T cells, as well as those miRNAs that may target mRNAs related to cell migration will allow for the understanding of the relevance of miRNA-mediated post-transcriptional control in the onset, development and progression of T cell lymphomas.

Description of Part II

Knowledge about the composition of the metabolic networks, which represent the set of chemical processes and reactions occurring in a cell, has been steadily improving over the last years, resulting in large collections of pathways concerning several species made available in public databases. This allowed the development of a large range of models and tools to study the structure of such networks in order to characterise their topological properties (Lacroix et al, IEEE/ACM Trans Comput Biol and Bioinf 2008, 5(4):594-617)^*. Such analyses can reveal important global organisational features common to the metabolic networks of many (if not all) species. These features can either be global, like patterns of node connectivity distribution (Ma & An-Ping, Bioinformatics 2003, 19:1423-30), or more local, like motifs and modules (Lacroix et al. IEEE/ACM Trans Comput Biol and Bioinf 2006, 3(4):360-368)^*.

The Portuguese group associated to this project has been developing research on the dynamic modelling of metabolic networks with promising results (Vilela et al., BMC Syst Biol 2009, 3:47; Vilela et al., BMC Syst Biol 2008, 2:35)^*. One of the objectives of this associate team project is to tighten the collaboration between the two sides, the French and the Portuguese, by putting together their respective expertise on the structural (static) and behavioural (dynamic) aspects of metabolic networks in the search for interconnections and mutual interferences between these two sides of a same coin. In this exploratory research effort, we already foresee two points that can yield relevant results and deserve further investigation. We briefly explain them in what follows.

Biochemical Systems Theory - BST (Voit, Cambridge University Press, 2000) involves the use of non-linear coupled differential equations, relating the fluxes in a pathway with the metabolite concentrations present in cells over time. In some general models, like in the so-called S-systems, the high number of parameters makes their estimation a very hard computational problem even for systems of modest size. Moreover, since the production of concentration measurements is relatively costly and time-consuming, the amount of training data for these models is limited. Hence, in many practical situations, the system results over-parameterised and over-fit to the experimental data. We argue that the information required to reduce the number of parameters or constrain their feasible intervals can come from the structural analysis of the network. For example, if we know that metabolic networks are modular, it may be inappropriate to consider all-against-all influences, as in the more general S-system formulation. We intend to investigate how structural information can be used in a systematic way to guide parameter estimation in dynamic metabolic models.

Optimisation tools (Torres & Voit, Cambridge University Press 2002) can be suitably applied in metabolic engineering, whose goal is to modify specific biochemical reactions through the manipulation of DNA. By identifying key control points it is possible to alter the expression of the enzymes catalysing specific reactions, thus altering the fluxes and/or concentrations, both transiently and at steady-state. The analysis of structural properties of metabolic networks, like the identification of precursor sets of target metabolites, that is, the minimal set of compounds sufficient to produce those targets (Cottret et al., Lect Notes Comput Sc/Lect N Bioinformat 2008, 5251:233-244)^*, can play an important role in the control engineering of metabolic networks by revealing potential sets of metabolites and fluxes that must be adjusted in order to achieve the desired behaviour of the network. On the opposite direction, the control manipulations on metabolic networks can suggest indirect connections between metabolites which, once scrutinised from a rigorous combinatorial perspective, can reveal subtle unanticipated structural constructions in those networks which could also be found elsewhere and help to cast light on analogous situations.

^* Work co-authored by a member of the team.

2. Présentation des partenaires (1 page environ par partenaire)
/ Partners presentation (1 page per partner)

INRIA BAMBOO team

Researchers from the INRIA BAMBOO team belong to the INRIA and to the University Claude Bernard (Lyon 1). The team has a background mainly in the design and analysis of algorithms as well as in combinatorics and statistics for biology, and in bioinformatics. It is strongly interdisciplinary, gathering together people from computer science, mathematics and biology (both theoretical and experimental).

Team members
Marie-France SAGOT, INRIA, responsible, CV is found below, computational biology/algorithmics/combinatorics, director of research
Christian Gautier, University Claude Bernard and BAMBOO, statistics/computational biology, full professor
Vincent Lacroix, University Claude Bernard and BAMBOO, computational biology/algorithmics/combinatorics, associate professor
Cristina Vieira-Heddi, University Claude Bernard and BAMBOO, experimental biology, associate professor
Augusto Vellozo, University Claude Bernard and BAMBOO, post-doc
Vicente Acuña, University Claude Bernard and BAMBOO, PhD student
Paulo Milreu, University Claude Bernard and BAMBOO, PhD student
Nuno Mendes, INESC-ID, Instituto Superior Técnico and University Claude Bernard and BAMBOO, PhD student

CV of Marie-France Sagot
Marie-France Sagot was born on April 21st, 1956, in São Paulo, Brazil; she obtained a Bachelor of Science degree in computer science in December 1991 at the University of São Paulo, Brazil, a PhD in Theoretical Computer Science in July 1996 at the University of Marne-la-Vallée, France and the Habilitation a Diriger les Recherches (HDR) in July 2000 from the same University. She was appointed associate researcher at the Pasteur Institute in Paris from 1997 to 2001, then associate researcher from 2001 to 2003 at the INRIA in Lyon, in the HELIX project-team. In 2003, she was promoted to director of research at the INRIA.

She has co-authored about 60 papers in journals, international conferences or as book chapters. Her current research interests concern design and the analysis of algorithms notably for biology.

She has co-founded one national and one international conference (resp. JOBIM and the European Conference on Computational Biology - ECCB) and (co-)chaired national and international conferences and workshops (JOBIM 2000 - Montpellier; JOBIM 2004 - Montreal, Canada; ECCB 2003 - Paris; CompBioNets 2004 - Recife, Brazil; CompBioNets 2005 - Lyon; BSB 2007 - Rio de Janeiro, Brazil; ISMB-ECCB 2009 - Stockholm). She has served, resp. serves, in the steering committee of JOBIM and ECCB. She is associate editor in Lecture Notes in BioInformatics, Journal of Discrete Algorithms, BMC Bioinformatics, BMC Algorithms for Molecular Biology and IEEE/ACM Transactions on Computational Biology and Bioinformatics of which she became the editor-in-chief starting from January 2009.

She is or has been the PI, for her team or general, of several national and international grants (Welcome Trust, Royal Society UK, Brazilian and Portuguese projects, ARCs INRIA, ACIs, ANRs of which two as sole funded team).

She founded and directed from 2004 to 2007 the first PhD Program on computational biology in Portugal, is a visiting researcher fellow at King's College, London (since 2002) and at the INESC-ID, Instituto Superior Técnico, Lisbon (since 2004).

She has served as committee member in the Comité National de la Recherche Scientifique CID 44 (2004-2008), has served in various researchers selection committees for the INRIA and is currently serving in the 01 (Maths) section and the CID 43 (interdisciplinary) of the CNRS.

KDBIO Team

Researchers from the KDBIO team belong to INESC-ID, Instituto Superior Técnico (IST), Faculdade de Ciências Médicas (FCM) and Faculdade de Ciências e Tecnologia (FCT). This team develops its activities in the areas of knowledge discovery, computational biology and bioinformatics. The objective is to integrate competences from researchers in diverse fields, including computer science, biology, statistics and control to address the challenges in computational biology and bioinformatics. The team research is cented in the areas of algorithms and complexity, machine learning and data mining, and databases and information systems.

Team members
Ana Teresa Freitas, INESC-ID/IST, responsible, CV is found below, computational biology/algorithmics/data mining, assistant professor (with tenure)
Arlindo Oliveira, INESC-ID/IST, algorithms/machine learning/data mining/computational biology/systems biology, full professor
Susana Vinga, INESC-ID/FCM, computational biology/statistics/systems biology, senior researcher at INESC-ID and invited assistant professor at FCM
Paulo Fonseca, INESC-ID, computational biology, senior researcher at INESC-ID
Luís Russo, INESC-ID/FCT UNL, algorithms/combinatorics, assistant professor at FCT and invited researcher at INESC-ID
Alexandra Carvalho, INESC-ID/IST, teaching assistant at IST, PhD student
Francisco Fernandes, INESC-ID, Instituto Superior Técnico, PhD student
Nuno Tenazinha, INESC-ID/ITQB-UNL, PhD student

CV of Ana Teresa Freitas
The coordinator of the Portuguese side is Prof. Ana Teresa Freitas, assistant professor (with tenure) at the Department of Electrical and Computer Engineering of Lisbon Technical University (Instituto Superior Técnico, IST). She is also a senior researcher at INESC-ID in the Knowledge Discovery and Bioinformatics group. She was born on June 23rd, 1967, in Madeira, Portugal. She got the B.S. and M.Sc. degree in Electrical and Computer Engineering from IST in 1990 and 1994, respectively. She got a PhD in the area of computer aided design with applications to dynamic systems modelling in 2002, from IST and her tenure from the same university in 2007.

She has co-authored about 30 papers in journals and international conferences in the areas of computational biology, bioinformatics and computer aided design. Her research interests are now centered in the areas of Computational Biology, Data Mining, Algorithms and Complexity.

The Knowledge Discovery and Bioinformatics group (KDBIO), headed by she and Arlindo Oliveira, is the most active and dynamic research group in his area, in Portugal, and integrates 6 PhDs, and several postdoctoral fellows and PhD students.

She is, or has been, PI of several Portuguese projects, two of them with participation of the INRIA BAMBOO team, and participates in two European projects, being PI for INESC-ID in one of these projects. She advised several MSc students and advises now two MSc students and four PhD students, one in co-supervision with Marie-France Sagot from the INRIA BAMBOO team. More information on her CV is available from her web page: http://kdbio.inesc-id.pt/~atf/

Brief historic of the collaboration

The BAMBOO and KDBIO teams have been collaborating for five years on several bioinformatics topics. They already have a number of publications in common, listed below.

[1] Nuno Mendes and Ana T. Freitas and Marie-France Sagot. Current tools for the identification of miRNA genes and their targets. Nucleic Acids Research, 8(37), pp. 2419-2433, May. 2009
[2] Alexandra M. Carvalho and Arlindo L. Oliveira and Marie-France Sagot, Efficient learning of Bayesian network classifiers: An extension to the TAN classifier, Proceedings of the 20th Australian Joint Conference on Artificial Intelligence, Dec. 2007 , pp. 16-25 , Springer-Verlag.
[3] Alexandra M. Carvalho and Ana T. Freitas and Arlindo L. Oliveira and Marie-France Sagot, An Efficient Algorithm for the Identification of Structured Motifs in DNA Promoter Sequences, IEEE Transactions on Computational Biology and Bioinformatics, 3(2), pp. 126-140, Apr. 2006, IEEE.
[4] Alexandra M. Carvalho and Ana T. Freitas and Arlindo L. Oliveira and Marie-France Sagot, A highly scalable algorithm for the extraction of cis-regulatory regions, Proceedings of the 3rd Asia Pacific Bioinformatics Conference, Jan. 2005 , pp. 273-282 , Imperial College Press.
[5] Alexandra M. Carvalho and Ana T. Freitas and Arlindo L. Oliveira and Marie-France Sagot, Efficient extraction of structured motifs using box links, Eleventh Symposium on String Processing and Information Retrieval, Nov. 2004 , pp. 267-268 , Springer.

Besides those common publications, the two teams share two PhD students: Alexandra Carvalho, in co-supervision between Arlindo Oliveira and Marie-France Sagot, and Nuno Mendes, between Ana Teresa Freitas and Marie-France Sagot. The first is in Lisbon while the second is in Lyon. The two teams have been frequently visiting one another over the years, either for research work or because Marie-France Sagot has participated in the thesis committee of PhD students of Arlindo Oliveira and vice versa. Another current member of the KDBIO, Susana Vinga, had in the past also twice paid a longer visit to the BAMBOO team (resp. of 2 and 1 months) when she was still preparing for the PhD. Finally, Paulo Fonseca that had, until recently, a Post-doc and a doctoral student position at BAMBOO team is currently a senior researcher of the KDBIO group in Lisbon.

The teams have furthermore had various projects in common, in particular Portuguese projects of which the last ones are: "ARN - Algorithms for the identification of genetic Regulatory Networks", funded by the FCT for the period from 2008 to 2010 (project reference: PTDC/EIA/67722/2006) and "microEGO - Did you ask for something small? The microRNAs power in a Eucalyptus tension world!", funded by the FCT for the period from 2010 to 2013 (project reference: PTDC/AGR-GPL/098179/2008).

3. Impact (1 page maximum)
/ Impact (maximum 1 page)

Part I - Algorithms and models for predicting small functional RNA motifs and their targets

This work on miRNAs involves the analysis of a vast amount of heterogeneous biological data, a great part of which will be generated in the context of this project. Efficient computational tools are needed to extract useful knowledge from it. The association between the teams aims precisely at addressing this point by bringing together people from several areas of computer and biological sciences. An external collaboration is also already established with Ana Tereza Vasconcelos from the LNCC, in Brazil. This researcher is responsible for the genome sequencing of Anopheles darlingi, a Brazilian National genome project. Her laboratory will provide the genome annotation and help in the data analysis. She will be also responsible for the expression profiling of the microRNAs in T cell lymphomas by high-throughput sequencing.

On the theoretical side, contributions are expected in the areas of text processing and classification. The search for regions obeying particular but flexible rules over large text corpora is a theoretical problem of interest per se, involving the design of efficient index structures and approximate search and inference algorithms. The posterior classification of candidate precursors and miRNA based into positive or negative instances is a machine learning task that comprises the development of sensitive filters and averaging schemas for combining them. The computation of likelihood and p-value scores associated to this classification is a problem that raises both statistical and computational issues.

On the biological side, little is known about which miRNAs are expressed in infectious diseases vector insects following parasite infection or in T cell lymphomas. Determining such microRNAs may represent an important contribution to the understanding of the physiopathology of these diseases, opening the way for the study of novel therapeutic targets. This work will be conducted in collaboration with Ana Tereza Vasconcelos and with Wilson Savino, head of the Laboratory on Thymus Research, Department of Immunology of the Oswaldo Cruz Institute in Brazil.

Part II - Exploring the interconnections between structure and dynamics of metabolic networks

A comprehensive and integrative view of structural and dynamic aspects of biological networks is the ultimate goal of systems biology. Hence the results obtained within this project are likely to have a relevant impact in the field and in its integration with control theory and dynamic systems. The development of mathematical and computational methods tailored for metabolic networks can be adapted to other areas of biology, biochemistry and genetics, or even beyond the realm of the life sciences.

Basic results are expected in the field of small-sample parameter estimation of non-linear systems. In the metabolic networks case, specialised ad-hoc information about the structure of specific pathways is used to tune parameter estimation. However the integration of structural information into parameter estimation in a systematic way remains an important open issue. Furthermore, the study of biochemical systems can provide new challenges to control theory and optimisation. Natural evolution has resulted in biological systems that are highly robust and fault-tolerant, which can be a source of inspiration for the innovative design of engineering components.

Finally, in addition to the scientific aspects discussed above, this project will serve the practical purpose of enlarging and solidifying the interaction between the French and Portuguese teams, as well as the external collaborators. It is our intention to promote the mobility of our PhD students and young researchers (some of which have just had their positions) and to let them assume active roles in an international project of a manageable size so as to develop even further their scientific competence and organisational skills. Also, by organising an open workshop, we intend to gather attention to our work and to the field, and to establish new research connections.

Description du programme scientifiquede travail (1 à 2 pages maximum)
/Description of the scientific work programme (maximum 1 to 2 pages)

The program for 2010 will be divided into two main parts, following the two main topics described in the scientific goals. The first part will concern small RNA motifs and targets identification and the second establishing methodological and biological links between the structural exploration of a metabolic network and its dynamic study. Work on the first part has already started in the context of the PhD of Nuno Mendes co-supervised by Ana Teresa Freitas and Marie-France Sagot, but is new in terms of its applications to the genomes of the mosquitoes Anopheles gambiae and Anopheles darlingi as well as the fruit fly, and specially, it is completely new as concerns the human data on T cell lymphomas. This application may require a revision / adaptation or extension of the method develop so far. Work on the second part is completely new to the collaboration, and therefore, the first few months of 2010 will most probably be spent in understanding what the other partner is doing, and in arriving at a common vocabulary and technical expertise.

Part I - Algorithms and models for predicting small functional RNA motifs and their targets

RNA genomics is still in the early stages of development. Most computational approaches developed so far make extensive use of evolutionary conservation either to predict miRNA genes or miRNA:target associations. The algorithmic approaches to this problem should be innovative with an emphasis on the application of combinatorial methods as well as machine learning techniques leading to the ab initio inference of small RNA regulatory signals. Statistical methods need also to be developed in order to calculate the statistical significance of the candidate motifs. This task will have three main objectives.

I.1 New computational method to predict small functional RNAs: The new method, which is already under development as part of the PhD work of Nuno Mendes, consists of a high-sensitivity low-specificity method to identify the largest possible set of candidate miRNA precursors in a genome of interest. The subsequent elimination of candidate miRNA genes is to be made by considering an extensible set of evidence-based criteria. The strength of this approach is its modularity and the ease with which one can modify one aspect of the pipeline to incorporate the rapidly changing knowledge on the field of miRNA biology. Furthermore, this framework can be used to analyse pre-miRNA candidates determined by a high-throughput sequencing technology and mapped back to the genome where precursor candidates can be extrapolated.

The evidence-based criteria that are already under development are the following ones: Adjusted Minimum Free Energy; Robustness of folding, Robustness with respect to context; Robustness with respect to mutations; Structure-based clustering of candidate stem-loops; Neighboring candidates and Annotation.

I.2 Comparison of the miRNA sequences with their homologs in different organisms: As many authors are now arguing that the identification of well-conserved and phylogenetically extensive miRNAs is reaching its saturation, a question that is becoming more important is whether non-conserved, presumably more exotic, miRNA precursors would be processed as such in different organisms that may have small yet important differences in their processing pathways. The elucidation of this question is crucial to methods which try to generalise from pre-miRNAs taken from several different species.

I.3 Target prediction: This area has received a new impetus with the recent proposal of a thermodynamic model incorporating target accessibility. However, seed matches still seem to serve as an important sieve to control false positives. Recent research has showed that at least some experimentally confirmed targets seem to violate the seed rule by including an unusual amount of mismatches or G:U pairs. New models for the background distribution of nucleotides in 3' UTR regions are needed and may lead to improvements in miRNA target predictions.

Part II - Exploring the interconnections between structure and dynamics of metabolic networks

The reverse engineering of biochemical pathways still constitutes a major challenge in systems biology. Lack of complete information about the system severally hampers model identification. On the other hand, insufficient experimental data might constitute an additional problem since it will affect parameter identifiability. For these reasons, the models obtained are usually over-parameterised, which causes convergence problems of the optimisation algorithms, in particular parameter estimation procedures.

II.1 The first topic to be addressed is to develop new tools to improve model identification by combining and integrating structural and dynamical information. In fact, knowledge about feasible metabolic networks in terms of topology, i.e., those with given structural properties, should be included when inferring the parameters of the dynamic counterpart. This is expected to improve the optimisation algorithms for parameter estimation, in terms of convergence and accuracy, since it dramatically reduces the search space. During this task, the systematisation of this process in an efficient way will be addressed.

II.2 The second topic is related to metabolic engineering and control. The study of the interdependencies between the metabolites can be analysed by developing algorithms in graphs. The static (structural) part has been already pursued (French team), and the optimisation of simple pathways using control strategies has also been initiated (Portuguese team). However, the systematic integration of these strategies is still at an early stage. During this project, algorithms based on control engineering will be developed for predicting sets of alterations on the network, both structural and dynamical through parameter changes, in order to optimise some target function. This can be the production of a given metabolite, the maximisation of growth/biomass, or the alteration of the dynamic properties of the system, for example, in order to induce bistability or convergence to a given steady-state. The integration of these methodologies is expected to bring new insights on the interdependencies between the metabolites involved.