Job Opportunities
5 Post-doc positions at PSB
The Department of Plant Systems Biology is recruiting 5 post-docs. For more information, click here:
http://www.psb.ugent.be/jobs/579-5-postdoctoral-positions
Open position: Statistician modeler
A 16 month position for a statistician modeller is available at the Research Unit on Improvement, Genetics and Physiology of Forest Trees (http://www.orleans.inra.fr/les_unites/ur_agpf) at INRA Orleans (France).
Starting date should be as soon as possible after October 1st 2010. The application deadline will be September 1st 2010.
Context:
Integrating molecular markers in breeding programs is a challenge in forest tree breeding. To optimize breeding strategies, the "Genetics" research team at UR AGPF develops Monte-Carlo simulation programs concerning the genetic evolution of populations subjected to artificial selection. The recruited statistician will be in charge of 1) adding new methods for genotypic evaluation to the simulation program on the basis of algorithms found in the scientific literature and testing their implementation with various sets of parameters 2) the development of a graphical interface to facilitate the use of the simulation tool by breeders. The position is opened in the framework of WP3 “Designing and implementation of improved/novel tree breeding strategies” of NovelTree European research project (http://www.noveltree.eu). The simulation tool resulting from this work will be used in priority to optimize the poplar breeding strategies managed by the Research Unit but will also be of interest for other forest tree species.
Job details
Adapt and implement in the existing simulation program statistical routines found in scientific literature or public repositories to predict breeding values using the BLUP (Best Linear Unbiased Prediction) methodology. Participate actively to the choice of the best fitted and performing algorithms to integrate genotypic data (molecular markers) in the mixed models: particularly, the generation of IBD (identity by descent) matrices at marker and QTL. Suggest improvements to the simulation program, especially in all parts concerning statistical treatment of the data, and implement the chosen improvements. Some parts of the software will be validated by comparison with existing tools (ASReml, Gibbs samplers for genetic mixed models, etc.). Conceive and build an interface facilitating the use of the simulation tool by breeders, in connection with potential users.
Working Environment
The successful candidate will work in the "Genetics" research team, under the co-responsibility of
Leopoldo Sanchez and Helene Muranty.
Education and training
Master degree in statistics
Required skills:
- Expert knowledge in at least one programming language (C ++ / FORTRAN / Java)
- Working knowledge of the Linux / UNIX system.
- Good scientific and technical English skills.
Gross salary:
1915 euros monthly
For more details and to apply for the post, please contact
H. Muranty (Helene.Muranty@orleans.inra.fr / 33 2 38 41 78 47) or
L. Sanchez (Leopoldo.Sanchez@orleans.inra.fr / 33 2 38 41 78 14)
INRA UR Amélioration, Génétique et Physiologie Forestières, 2163 avenue de la Pomme de Pin -
CS40001 - Ardon, 45075 ORLEANS CEDEX 2, FRANCE
Open position: Computer Software Engineer
A 16 month position for a computer software engineer is available at the Research Unit on Improvement, Genetics and Physiology of Forest Trees at INRA Orleans (France).
Computer programs were developed in FORTRAN (95) to simulate the genetic evolution of populations subjected to artificial selection. The recruited engineer will be in charge of 1) adding new methods for genotypic evaluation to the simulation program on the basis of algorithms provided by the scientific team and testing their implementation with various sets of parameters 2) the development of a graphical interface to facilitate the use of the simulation tool by breeders. The position is opened in the framework of WP3 “Designing and implementation of improved/novel tree breeding strategies” of NovelTree European research project (http://www.noveltree.eu/index.php).
The simulation tool resulting from this work will be used in priority to optimize the poplar breeding strategies managed by the Research Unit but will also be of interest for other forest tree species.
Starting date should be as soon as possible after March 1st 2010. The application deadline will be February 20 2010.
Job details
Collect, adapt and optimize existing statistical routines found in scientific literature or public repositories to implement them in the simulation program. These routines concern the BLUP (Best Linear Unbiaised Prediction) methodology, which uses genetic relationships from a given pedigree to predict individual breeding values in a breeding population. In order to use genotypic data to predict breeding values, write routines and test them by simulation in the framework of the existing simulation program. This will comprise the generation of IBD (identity By descent) matrices at marker and QTL positions and their integration in the BLUP evaluation. Some parts of the software will be validated by comparison with commercial tools (ASReml, Gibbs samplers for genetic mixed models, etc.). Conceive an interface facilitating the use of the simulation tool by breeders, in connection with potential users, and build it.
Working Environment
The successful candidate will work in the "Genetics" research team, under the co-responsibility of Leopoldo Sanchez and Helene Muranty.
Education and training
Master degree in computer sciences
Required skills
- Familiarity with statistics and matrix algebra
- Expert knowledge in at least one programming language (C ++ / FORTRAN / Java)
- Working knowledge of the Linux / UNIX system.
- Good scientific and technical English skills.
Gross salary: 1915 euros monthly
For more details and to apply for the post, please contact
H. Muranty (Helene.Muranty –at- orleans.inra.fr) or
L. Sanchez (Leopoldo.Sanchez –at- orleans.inra.fr)
INRA UR Amélioration, Génétique et Physiologie Forestières, 2163 avenue de la Pomme de Pin -
CS40001 - Ardon, 45075 ORLEANS CEDEX 2, FRANCE
Open position: Post-doctoral position
Subject
18 months post-doctoral position at INRA Orléans, France
Host institution
AGPF Unit - INRA Orléans (FRANCE)
Key words:
Bayesian statistics, MCMC, marker-assisted BLUP, computer simulation, dominance, breeding metapopulation structure
Applications closing date:
November 15th, 2009.
Application forms will be provided upon request writing at the following addresses:
leopoldo.sanchez@orleans.inra.fr
catherine.bastien@orleans.inra.fr
The following postdoctoral project is part of NOVELTREE large-scale integrating project (May 2008-April 2012), which aims at developing improved breeding strategies in the context of forest tree breeding programs.
The present postdoctoral project involves two alternative sub-projects. Candidates must have a strong background in at least one of the two sub-projects (see below), in which the successful candidate will focus her/his research efforts, with the possibility to be involved in the other sub-project depending on the capabilities/interest of the candidate.
Objective(s)
Two sub-projects are proposed whose respective objectives are:
- Implementation of bayesian and MCMC methods for statistical inferences with multivariate phenotypic and genotypic records in genetic evaluation for poplarbreeding (denoted hereafter sub-project A).
- Optimization of breeding meta-population structures for genetic improvement purposes by using allele-based simulations under non-additive assumptions (denoted hereafter sub-project B).
These two objectives are related to each other in the context of a breeding program. While the first objective concerns the preliminary phase of genetic evaluation of candidates to selection by using best suited statistical tools, the second objective concerns the subsequent phase of weighting selection decisions observing other breeding constraints that are aimed at preserving a given diversity structure by using optimization and simulation tools.
Context
Concerning the sub-project A, one important breakthrough in computational statistics has been the MCMC (Markov chain Monte Carlo) with a rapid adaptation to the needs of quantitative genetics, particularly in the genetic evaluation in animal breeding. One of the benefits leading this expansion is the huge flexibility of the models that can be implemented under MCMC, especially from a Bayesian perspective, which is in contrast with the strictness of standard methods of statistical analysis based on infinite size inferences. Few are still the examples with MCMC applied to forest tree breeding, despite the fact that some circumstances in these programs appear particularly favorable to the application of these techniques. Often with designed selection experiments involving different sources of phenotypic and genotypic records, an elaborate multivariate structure with a large number of nuisance parameters is needed for fitting. Conversely, there is a certain scarcity of resources available being used in the test. These two circumstances dictate much uncertainty about the parameters to be inferred. Arguably, inference based on asymptotic theory is not expected to be satisfactory under these limiting circumstances. Other complications arise when data available is not the result of random sampling and selection has already modified the genetic composition of the population from which data is to be obtained. An alternative to overcome these difficulties is the Bayesian approach. MCMC is a very important computational tool in Bayesian statistics, as it allows inferences from complex distributions where analytical or numerical integration is not available.
Regarding the sub-project B, breeding populations comprise often several interconnected breeding units, each one serving a particular breeding purpose, being the result of a given geographical origin distinct from the other units or being a different species able to interbreed with the other species in the breeding population. This latter case is the one of the poplar breeding program, where several species are intercrossed to benefit from heterotic responses generally for growth traits in the resulting hybrid offspring. Tools based on optimization have been developed to manage selection decisions and mating regimes within a single breeding unit i.e. a unique breeding population. These tools share the principle of maximizing genetic gain obtained from selection while constraining the loss of diversity that might result from that gain. Some of these tools target the genetic contribution of individuals to subsequent generations as decision variables, while others target directly matings between candidates. All of them have been shown as beneficial in the long-term management of breeding resources over standard unconstrained methods in simulation studies, in a series of lab experiments and in a few advanced animal breeding programs. However, their benefit has not been evaluated in the context of a breeding meta-population program comprising multiple interbreeding units like the one represented by the poplar program. Additionally, most of the models used to evaluate the benefit of these methods assume additive inheritance. However, non-additive variation might play an important role in crossbreeding, since this is one of the most effective ways to exploit heterosis. The use of optimum selection and mating procedures has been suggested in the context of exploiting non-additive variation.
Methodologies
Sub-project A: Poplar breeding program provides multivariate phenotypic and genotypic data, the latter comprising numerous polymorphisms at genetic markers of Quantitative Trait Loci for several important traits concerning resistance to rust, traits related to growth and phenology.
- Bayesian inference of genetic parameters will be implemented, involving MCMC in one of its various forms like Gibbs sampler (Metropolis-Hastings algorithm) or the reversible jump MCMC when considering unfixed QTL effects. The aim is to implement best suited routines of these methods to provide (co)variance components estimations for the various traits of growth and rust resistance in the poplar breeding population in France, as well as their corresponding posterior distributions for further inferences.
- Second, conditionally on these variance components estimates, best linear unbiased predictions of the genetic effects of polygenic background and of unlinked QTLs for each considered trait will be obtained. The ultimate objective of the analysis is to provide a genetic evaluation based on genotypic and phenotypic data for all candidates in the poplar breeding population, as well as their posterior distributions.
- These procedures will be first implemented on an allele-based simulator mimicking available poplar data in order to test different evaluation scenarios and economic weights. Ultimately, the analytical procedure will be applied to real poplar data with a standalone bayesian MCMC program.
Sub-project B: The current poplar breeding structure will provide a template for simulating a breeding metapopulation. Comprehensive datasets from poplar evaluation program on (co)variance structures between selected traits and genetic mapping and QTL information will feed the underlying gene model of the simulation. The following methodologies will be implemented into a single simulation package:
- Development of an allele-based simulation model for quantitative trait loci and neutral background, where loci layout will be implemented by genetic maps and matrices of linkage disequilibrium between markers and QTLs, and where various models of dominance for the QTLs and polygenic background will be tested.
- Implementation of optimization algorithms to the previous simulator that will serve to obtain a given level of genetic structure between the breeding units of the metapopulation. This procedure will generate genetically distinct populations within the metapopulation based on genetic differentiation estimates like those of Wright and derivatives, and by generating LD between populations.
- Implementation of optimum selection and mating procedures into the simulator. These procedures will perform a maximization of genetic gain under constraints on diversity either at the level of selected parents or at the level of selected matings.
- Several scenarios will be tested mimicking the possibilities reasonably at hand for the poplar breeding program, with the aim of evaluating the benefits of optimum selection and mating procedures in exploiting non-additive quantitative trait loci variation over standard selection procedures. Since optimum selection procedures are expected to maximize effective population size for a given level of response over standard unconstrained selection, those reductions in population size under optimum selection for poplar breeding that guarantee similar levels of diversity and response will be identified.
Candidates
Candidates must have a strong background in statistics and/or quantitative genetics, with experience in genetic evaluation using frequentist and/or bayesian approaches. They should be able to understand and implement these methodologies. Knowledge in programming in one or several languages (C, FORTRAN, etc) will be needed. Some experience in forest tree breeding will be valuable.
Candidate must provide before the closing date of the call a Curriculum vitae and a short letter detailing his/her expected contributions to the project to the following addresses: leopoldo.sanchez@orleans.inra.fr, catherine.bastien@orleans.inra.fr
Host institution
Although the project involves general theoretical approaches, great emphasis will be given to place the project in a forest tree context, and to feed on the breeding research activities carried by the host unit (AGPF Unit - INRA Orléans). This project will be at the core of a larger research program where the main objective is to develop improved breeding strategies for several key species in France. AGPF has already an important experience in the development of mathematical models in the context of quantitative and population genetics. The research involved in this project is part of a larger major EU project which will foster multilateral collaborations around the subject of optimal procedures for forest tree breeding.
