The team is interested in emerging statistical issues in public health and clinical research as access to and use of large-scale data increases. Drug safety assessment and predictive genomics are our two main areas of methodological development, both of which involve high-dimensional issues, notably with machine learning approaches.
Our first axis is devoted to the evaluation of drugs in population, with methodological developments in pharmacoepidemiology and pharmacovigilance. We have a particular interest in vaccines and drugs taken during pregnancy. Our recent orientations also involve the exploitation of medico-administrative databases such as the National Health Data System (SNDS) in France. In pharmacoepidemiology, our objective is to develop statistical methods for the exploitation of SNDS data taking into account measured and unmeasured confounding. In addition, we are working on methods for estimating public health impact: the benefit-risk ratio of vaccines and attributable risk. In pharmacovigilance, we are developing methods for automatic signal detection aimed at exploiting the large databases of spontaneous reports and, more recently, of the SNDS. These methods are based on the statistical fields of high-dimensional variable selection and causality.
Our second axis carried by the Genostat group is devoted to predictive genomics and epidemiological genetics. The objectives are among others: (i) the development of integrative strategies that use various sources of "internal" (e.g. clinical samples) or "external" (e.g. databases) information, in particular internal multilayered information from genomics, transcriptomics and proteomics for prediction; (ii) the development of methods for association testing with rare variants, in particular for exome sequencing data.