Vacant jobs
A 2-YEAR POSTDOC POSITION
AVAILABLE
AT GENOPOLE-University of Evry-FRANCE in early
2010
in MACHINE LEARNING and REVERSE-MODELING OF
SIGNALING and GENE REGULATORY NETWORKS
The Machine Learning, Modeling and Data Integration group at laboratoryIBISC at
University of Evry -
Genopole (30km Paris, France) is seeking
postdoctoral candidates for a two-year position with a strong background in statistical learning
(graphical models and kernel methods) and an interest for interdisciplinary research aimed towards
reverse-modeling of biological networks.
The successful applicant will participate to an ANR project called
ODESSA* (Ordinary Differential Equations and State-Space models for signAlling and regulatory
networks inference) that will begin on January 1rst, 2010, in collaboration with the Laboratoire de
Biologie Moléculaire de la Cellule (Ecole Normale Supérieure de Lyon, France). He/she will
investigate new algorithms for structure and module learning in nonlinear state-space models. A
sound knowledge in graphical models (dynamical models, mixture models, latent variable models) and
kernel-based approaches is asked. The targeted application is the identification of signaling and
gene regulatory networks involved in the cellular response to retinoic acid from large scale data
(next generation sequencers). Experience in large scale data-mining of post-genomic data will be
appreciated but is not required.
The candidate will have a Ph.D. in Computer Science,
Mathematics or Bioinformatics with a strong publication record.
Applications for this role should include a statement of research
experience and interests, a CV, two reference letter(s) from appropriate academic sources and
the two best publications.
Please send your applications in electronic form to Prof. Florence d’A
lché-Buc : florence (dot) dalche (at) ibisc (dot) fr
Website of the team: http://amisbio.ibisc.fr
The position can start in 2010 and if possible not later than September
2010.
The deadline for application is July 10, 2010
ODESSA is described here:
ODESSA : Ordinary Differential
Equation and State-Space models for signalling and gene regulatory networks
inference
ANR
Project
Soon (January
2010)
www.odessa-project.fr
Partners:
- IBISC FRE CNRS 3190, Univeristé d’Evry-Val d’Essonne,
Genopole, Evry,France
- Institut de Génomique Fonctionnelle de Lyon
(IGFL)
Coordination:
Prof. Florence d’Alché-Buc :
florence.dalche@ibisc.fr
, amisbio.ibisc.fr
Keywords
: signalling networks, gene regulation networks, next
generation sequencers, network inference, feature selection, module extraction, constraints,
parametric and non parametric ODEs, state-space models, structure learning, probabilistic graphical
model, Bayesian inference, Population-based methods (Monte-Carlo, EDA, Gibbs sampling),
kernels
Abstract
Complex signalling and regulation mechanisms at work in
the cell involve different kinds of molecular components that interact through time. Identifying
these interactions and understanding how the cell implements a response to a given input signal are
undoubtedly the main challenge of systems biology. To address this issue is one of the most
powerful way to build a mathematical model of the complex dynamical underlying system and to study
how it gives support to biological hypotheses. Many modellers have significantly contributed to
this research current by exploiting different modelling frameworks like Ordinary Differential
Equations, Boolean and discrete-value networks with an analysis of the dynamics derived from
bifurcation theory or formal test methods.
Meanwhile several researchers ask another question: how
can we determine automatically the parameters and the structure of the dynamical system under
observation ? While for a long time, this goal was supposed to be out of reach, the spread-out of
new high-throughput technologies consecutively to the emergence of sequencing tools has completely
changed the game: it is now possible to measure gene expression in a given tissue in a given organ
of some organisms in given conditions and therefore, the dynamical system underlying the
interactions can be partially observed. Therefore, reverse-modelling becomes in principle feasible
although many technical problems have been raised these last years: non identifiability of some
parameters, limited sample size compared to the length of available time series, difficulty of
structure estimation in front of noisy and incomplete data.
This project, exploiting two concrete problems of
reverse-modelling, is mainly focused on these technical issues.
First biological problem:
The first biological issue is related to the analysis
of genetic networks controlled by transcription factors. We choose to focus our study on the
analyses of gene network modulated by Retinoic Acid Receptors (RARs), which are members of the
nuclear receptor (NR) superfamily. This family of ligand modulated transcription factors whose
study has been instrumental in the understanding of transcription regulation mechanisms, gathers
major transcription factors, sharing important structural and functional properties, and implicated
in development, in adult physiology as well as in various pathologies.
Second biological problem:
The second biological issue is related to the study of
oscillating biological systems and more specifically to the circadian clock. In the recent years,
lots of attention was put on the elucidation of the molecular mechanisms supporting the generation
of circadian oscillations in living organisms. Within this well-defined frame, we choose to focus
our study on the modulation and cellular integration of the circadian oscillations with a
particular emphasis on the role of the nuclear receptor rev-erb
a
.
Methods :
Starting from the wide class of Ordinary Differential
Equations and given the fact that the signalling network (resp. regulatory) may be not fully
observed, we propose to improve two frameworks for learning parameters and structure: the first one
is based on the idea that the observed trajectory can be approximated using a non parametric or a
semi-parametric model and that the derivative of the proxy can then be used in the cost function to
be minimized. The second one encapsulates the ODE into a discrete time probabilistic model such as
a state-space model in order to deal with hidden variables. The former framework opens the door to
the incorporation of qualitative constraints on the nature of the observed trajectory whereas the
latter framework benefits from the flexibility of graphical models and their dedicated leaning
approaches. Among issues related to the proposed biological problems, the project will try to
address the following ones:
-
feature selection, network inference from large scale
data (next generation sequencer), subnetwork extraction
-
incorporation of various constraints
in a Bayesian or frequentist framework: dynamics, biological function
etc…
-
new nonparametric models for state-space models and
differential equations
-
statistical test of models
Some recent publications related to the
project
[d’Alché-Buc & Brunel 2009] F. d’Alché-Buc, N., J.-B., Brunel. Estimation of
parametric nonlinear ODEs for biological networks identification, to appear as a chapter in
Computational Systems Biology: Data Driven Inference of Parameters and Network Structure, Neil
Lawrence, Magnus Rattray, Guido Sanguinetti,Mark Girolami, MIT Press, to appear in late
2009.
[d’Alché-Buc & Wehenkel 2008] F. d’Alché-Buc, L. Wehenkel, Machine
Learning in Systems Biology, special issue BMC proceedings, December 2008.
[Auliac et al. 2008] C. A
uliac, V. Frouin, X. Gidrol, F. d’Alché-Buc,
Evolutionary Approaches for the Reverse-Engineering of
Gene Regulatory Networks: a study on a biologically realistic Dataset, BMC Bioinformatics 2008,
9:91.
[Bouchet et al 2007]
P. Bouchet, N. Brunel, F. d'Alché-Buc, Parameter Estimation of ODEs
using Support Vector Regression and Qualitative constraints, PASCAL Workshop, September 18
th
2007, Glasgow
(communication).
[Brunel 2008] N.,J.-B.,Brunel, Parameter estimation of ODEs via
nonparametric estimators, Electronic Journal of Statistics,
Vol. 2 (2008) 1242–1267.
[Cai et al. 2008] Cai W., Rambaud J., Teboul M., Masse I.,
Benoit G., Gustafsson J.A., Delaunay F., Laudet V., and Pongratz I. Expression levels of ERβ
are modulated by components of the molecular clock.
Mol Cell Biol: 28(2), 784-793. 2008.
[Geurts et al. 2007] P. Geurts, N. Touleimat, M. Dutreix, F.
d'Alché-Buc, Inferring biological networks with output kernel trees, BMC Bioinformatics, 8(Suppl
2):S4, May 3, 2007.
[Geurts et al. 2007] P. Geurts, L. Wehenkel, F. d'Alché-Buc, Gradient
Boosting for Kernelized
Output Spaces, Zoubin Ghahramani (Eds.), in Proceedings of theTwenty-Fourth International
Conference on Machine Learning (ICML 2007),Corvallis, Oregon, USA, June 20-24, 2007,
227:289-296.
[Perrin et al. 2003] B.-E. Perrin, L. Ralaivola,A. Mazurie, S. Bottani,
J. Mallet, F. d'Alché-Buc, Inference of gene regulatory network with Dynamic Bayesian Network,
Bioinformatics (Oxford Press), vol. 19, pi38-49, (2003).
[Quach et al. 2007] M. Quach, N. Brunel, F. d'Alché-Buc,
Estimating parameters and hidden variables in nonlinear state-space
models based on ODEs for biological networks inference.
Bioinformatics, 23, pp. 3209–3216, 2007.
[Rambaud et al. 2009] Rambaud J., Triqueneaux G., Masse I., Staels B., Laudet V.
and
Benoit G. Rev-erbα2 mRNA encodes a stable protein likely to play a role in circadian clock
regulation.
Mol Endocrinol. In press. 2009
[Veber et al. 2008] P. Veber, C. Guziolowski, M. Le Borgne, O.
Radulescu, and A. Siegel. Inferring the role of transcription factors in regulatory networks. BMC
Bioinformatics, 9 :228, 2008.
|
Accessible Version
Imprint