DYLISS : Dynamics, Inference and Logics for bIological Systems and Sequences
DYLISS is a bioinformatics team focusing on the identification and characterization of the groups of genetic actors that control the phenotypic response of individuals or species to their environment.
The data revolution in life sciences combined with machine-learning techniques successfully highlight the top-ranking entities in a dataset. However, the biological interest lies in in the explanation of the ranking, and more precisely in the identification of the biological processes leading to the observed phenotypes. This requires to take into account the existing knowledge about the chains of biological compounds as well as their regulators.
The challenges are:
- integrating large heterogeneous complementary datasets and knowledge bases,
- extract explanation-supporting models of the observations compatible with the existing knowledge, and
- explore and analyze the exhaustive family of the consistent models.
DYLISS develops knowledge-based data analysis and reasoning methods spanning three axis:
- develop data structuration and integration methods to unify datasets and knowledge bases into knowledge graphs. This is supported by Semantic Web technologies and resources from the Linked Open Data initiative.
- take advantage of the structured data to extract families of models of the processes underlying the observations. This is achieved with a combination of learning methods.
- assist domain experts for exploring and analyzing the family of models.