Représentation des connaissances

Dyliss is a research team in bioinformatics. We focus on sequence analysis and systems biology. We use qualitative formal systems to characterize genetic actors from non model species, such as algae or mining baceria, that control phenotypic answers when challenged by their environment.

• Methods: constraint logic programming, symbolic dynamics, machine learning, formal systems.
• Expertise: functional characterization, non-model species, multi-scale integration.
• Application domains: marine biology, micro-environmental biology...

The main computational challenge is to face lacks and incompleteness in both expert knowledge and experimental observations. Our strategy relies on three main points.

• We focus on a targeted functional characterization rather than on a complete understanding of the species.
• We rely on knowledge rather than on amounts of experimentations.
• We use model-species to validate methods.

We use qualitative formal systems for knowledge acquisition and integration. All our methods aim at identifying the space of all models that are consistent with both knowledge and observations. Then we provide tools to navigate in this space in order to investigate which properties are shared by a large proportion of the space.

The research themes of the Orpailleur project-team hold mainly on the design of intelligent systems: knowledge systems, information systems, data mining systems (or knowledge discovery systems). These intelligent systems have various forms and may be used in a number of applicative contexts such as biology, chemistry, space, time, and on the Web.