Section: Research Program

Scalable methods to query data heterogenity

Confronted to large and complex data sets (raw data are associated with graphs depicting explicit or implicit links and correlations) almost all scientific fields have been impacted by the big data issueespecially genomics and astronomy  [47]. In our opinion, life sciences cumulates several features that are very specific and prevent the direct application of big data strategies that proved successful in other domains such as experimental physics: the existence of several scales of granularity from microscopic to macroscopic and the associated issue of dependency propagation, datasets incompleteness and uncertainty including highly heterogeneous responses to a perturbation from one sample to another, and highly fragmented sources of information that lacks interoperability [42]. To explore this research field, we use techniques from symbolic data mining (Semantic Web technologies, symbolic clustering, constraint satisfaction and grammatical modelling) to take into account those life science features in the analysis of biological data.

Research topics

Facilitating data integration and querying The quantity and inner complexity of life science data require semantically-rich analysis methods. A major challenge is then to combine data (from local project as well as from reference databases) and symbolic knowledge seamlessly. Semantic Web technologies provide a relevant framework, as demonstrated by the success of Linked (Open) Data  [34]. However, life science end users (1) find it difficult to learn the languages for representing and querying Semantic Web data, and consequently (2) miss the possibility they had to interact with their tabulated data (even when doing so was exceedingly slow and tedious). Our first objective in this axis is to develop accurate abstractions of datasets or knowledge repositories to facilitate their exploration with RDF-based technologies.

Scalability of semantic web queries. A bottleneck in data querying is given by the performance of federated SPARQL queries, which must be improved by several orders of magnitude to allow current massive data to be analyzed. In this direction, our research program focuses the combination of linked data fragments  [48], query properties and dataset structure for decomposing federated SPARQL queries.

Building and compressing static maps of interacting compounds A final approach to handle heterogeneity is to gather multi-scale data knowledge into functional static map of biological models that can be analyzed and/or compressed. This requires to linking genomics, metabolomics, expression data and protein measurement of several phenotypes into unified frameworks. In this direction, our main goal is to develop families of constraints, inspired by symbolic dynamical systems, to link datasets together. We currently focus on health (personalized medicine) and environmental (role of non-coding regulations, compression) datasets.

Associated software tools

AskOmics platform AskOmics is an integration and interrogation software for linked biological data based on semantic web technologies [url]. AskOmics aims at bridging the gap between end user data and the Linked (Open) Data cloud (LOD cloud). It allows heterogeneous bioinformatics data (formatted as tabular files or directly in RDF) to be loaded into a Triple Store system using a user-friendly web interface. It helps end users to (1) take advantage of the information readily available in the LOD cloud for analyzing there own data and (2) contribute back to the linked data by representing their data and the associated metadata in the proper format as well as by linking them to other resources. An originality is the graphical interface that allows any dataset to be integrated in a local RDF datawarehouse and SPARQL query to be built transparently and iteratively by a non-expert user.

FinGoc-tools The FinGoc tools allow filtering interaction networks with graph-based optimization criteria in order to elucidate the main regulators of an observed phenotype. The main added-value of these tools is to make explicit the criteria used to highlight the role of the main regulators.(1) The KeyRegulatorFinder package searches key regulators of lists of molecules (like metabolites, enzymes or genes) by taking advantage of knowledge databases in cell metabolism and signaling [package]. (2) The PowerGrasp python package implements graph compression methods oriented toward visualization, and based on power graph analysis [package]. (3) The iggy package enables the repairing of an interaction graph with respect to expression data. [Python package]