Project-Team : symbiose

French version

Team leader : Jacques Nicolas

Symbiose is a research project in the field of bioinformatics, whose general interest concerns the modelling of genomic and post-genomic data. Our goal is to assist the molecular biologist in the tasks of formulating his/her knowledge, reasoning on it and  discovering new relations in experimental data. Our research specificities include our interest in large scale studies (genomes or proteomes) and in filtering in data banks  sequences matching complex models.

Symbiose follows three main research axes:

• Analysis of sequences with formal languages: This track concerns the search for relevant (e. g. functional)  structures in macromolecules with the intent either to model specific spatial structures (secondary structures, disulfid bounds...) or general biological mechanisms (transposition, frameshift,...). We tackle these problems in the framework of language theory. Our first objective is to offer the biologist some means to build expressive models on macromolecules. We propose towards this goal a graphical environment, a string-based language and a calculus on suffix trees. A second objective is to automatically produce from two sets of sequences grammatical models able to discriminate one set from the other. We develop for this purpose researches in the field of grammatical inference, either on finite automata or higher level grammars.
  

• Combinatorial optimization and use of parallelism:  this axis set the issue of speeding up complex computations in  genomic aplications (comparison, intensive search for complex patterns, structure prediction...). We propose two types of studies in order to overcome current limitations: (1) reconfigurable architectures based on FPGA to speed-up filtering and discrete optimization techniques for a fine modelling of time-consuming calculations (particularly, integer programming for structure prediction). (2) Implementation of parallel indexes and use of fast large memories with the aim to seed up data access. We are also interested in GRID computing.

• Analysis and identification of dynamic systems:  We are interested in identifying genes with a noticeable effect on given metabolic and signalling pathways interacting with a genetic network. Our approach is based on the synthesis of interaction knowledge within a qualitative model and its derivation into discrete or differential models. Our aim is to be able, from the model of a particular biological system, to produce explanations of observations and predictions on its behaviour.