Information Sciences to support Personalized Medicine Team.

New technologies inspired are transforming biomedical research from a laboratory science to an information science. We need new approaches to making sense of the data we generate. Understanding disease is now being able to collect, manage, analyze, and interpret the data. The team project focuses on new methodologies that combine statistical and computer science methods along three axes :

Secondary use of patient clinical data: Re-use of routine care data has been made possible thanks to the clinical information systems. We have developed joint projects with the Hôpital Européen Georges Pompidou (HEGP), is a 800-acute-bed academic hospital located in Paris, that adopted a clinical information system in 2000. The system is not only used at every step of patient care, but also produces a vast repository of disease and treatment data –the clinical data warehouse– that can be mined for research. For example, secondary use of these data can be made for Phenome-Wide Association Studies (PheWAS) to investigate whether genetic polymorphisms associated with a phenotype are also associated with other diagnoses. We have conducted in collaboration with INSERM UMRS-775 the first PheWAS ever performed in France. In this PheWAS, we used HEGP EHR data for phenotypic information and we developed new methods that combine ICD-10 codes and biological test results for phenotypic information, and use a quantitative trait as the selection criterion. We tested our approach on TPMT activity in patients treated by thiopurine drugs, and demonstrated that such an approach allows for the identification of subgroups of patients who require personalized clinical and therapeutic management.

Integration of multiple heterogeneous biolomediical sources. Omics information has to be associated with comprehensive clinical information such as clinical manifestations, images, laboratory tests, pathological findings, and drug histories, which requires sound and unified phenotype annotation of genes, gene products and biological specimens, as well as phenotype enrichment. We rely on novel integrative platforms such as tranSMART. The objective is to refine research hypotheses by investigating correlations between genetic and phenotypic data, and assessing their analytical results in the context of published literature, pathways or publicly available biomedical database.

New methodologies for dose finding. Prospective individualization of drug dosing may prevent or minimize dose-limiting toxicity and allow patients to achieve the maximal treatment benefit by improving their ability to complete therapy. Early-phase clinical trials will benefit from exploitation of data to design smaller, shorter, and individualized clinical trials. Most of evaluated agents in clinical trials are not of “first in man” type; they were already evaluated and used in some indication associated with one or several administration modes. We are investigating methods reusing knowledge bases and data warehouses to extrapolate from existing information and design safer trials. We will take into consideration not only genetic biomarkers but also other patient characteristics such as childhood and comorbidities.