Section: Overall Objectives

Overall objectives

The research domain of the Dyliss team is bioinformatics and systems biology. Our main goal in biology is to characterize groups of genetic actors that control the response of living species capable of facing extreme environments. Unlike model species, a limited prior-knowledge is available for these organisms together with a small range of experimental studies (culture conditions, genetic transformations). To overcome these limitations, the team explores methods in the field of formal systems (knowledge representation, constraints programming) to take into account information on physiological responses of the studied species under various constraints (knowledge representation, constraint programming, multi-scale analysis of dynamical systems) as well as genetic information from their long-distant cousins (machine learning).

The challenge to face is thus incompleteness: limited range of physiological or genetic known perturbations together with an incomplete knowledge of living mechanisms involved. We favor the construction and study of a "space of feasible models or hypotheses" including known constraints and facts on a living system rather than searching for a single optimized model. We develop methods allowing a precise investigation of this space of hypotheses. Therefore, the biologist will be in position of developing experimental strategies to progressively shrink the space of hypotheses and gain in the understanding of the system. This refinement approach is particularly suited to non-model organisms, which have specific and little known survival mechanisms. It is also required in the framework of an increasing automation of experimentations in biology, which needs to better formalize the experimentation loop.

From the bioinformatics aspect, the main challenge is to transfer genome-level information available in well-annotated organisms on their distant relatives. To that matter, we develop methods within the context of formal systems to identify and formalize the genomic specificities of target species mainly observed at the physiological level. Our main purpose is to combine in a suitable way machine learning, logical constraints and dynamical systems techniques to get a combinatorial representation of the space of admissible models for families of genomic products implied in the living system response. The steps of the analysis are to (i) formalize and integrate in a set of logic constraints the genetic information and the physiological responses; (ii) investigate the space of admissible models and exhibit its structure and main features; (iii) identify corresponding genomic products within sequences.

We target applications in marine biology and environmental microbiology, that is, organisms with a good long-term biotechnological potential but requiring prior intensive in-silico studies to fully exploit their specificities. We focus on unicellular and pluricellular organisms with a relatively simple development but very specific physiological capabilities. Existing long-term partnerships with biological labs give strong support to this choice: in marine biology, we collaborate closely with the Station biologique de Roscoff (Idealg, Investissement avenir "Bioressources et Biotechnologies") whereas in environmental microbiology we collaborate both with the CRG in Chile in the framework of the Ciric Chilean inria center (Ciric-Omics) and with laboratories in Rennes (Inra).