Keynotes

We are proud to announce that Aad van der Vaart (Delft Institute of Applied Mathematics) and Marc Chadeau-Hyam (School of Public Health, Imperial College London) will respectively deliver the opening and closing keynote address. 

Opening Keynote address: Sensitivity Analysis in Causal Analysis

Aad van der Vaart (Delft Institute of Applied Mathematics, TU Delft)

Causal inference is based on the assumption of “conditional exchangeability”. This is not verifiable based on the data when using nonparametric modelling. A “sensititvity analysis” considers the effect of deviations from the assumption. In a Bayesian framework we could put a prior on the size and structure of the deviation and obtain an ordinary posterior. We review possible approaches and present some results on one approach, comparing different ways of nonparametric modelling. (Based on joint work with Bart Eggen, Stéphanie van der Pas and Chris van Vliet).

Closing Keynote address: Exposome Analytics: methodological developments and needs

Marc Chadeau-Hyam (School of Public Health, Imperial College London)

The Exposome concept has been developed as a necessary complement to the genome to better understand the determinants of health and of the risk of chronic diseases. The external exposome combines a large range of external stressors (i.e. non-genetic) factors potentially impacting human health from conception onwards. These external exposures (i) are heterogeneous in nature, scale, and variability, (ii) feature complex correlation patterns and (iii) may operate as mixtures. The internal exposome can be defined as the way these exposures are embodied and its exploration relies on the screening and integration of high-resolution molecular data. While methods for omics data analyses are established, their application in an exposome context is raising specific methodological challenges including the analysis of complex and correlated exposures. Furthermore, the isolated exploration of an omic profile offers the possibility to capture stressor-induced biological/biochemical alterations, potentially impacting individual risk profiles, but this may only yield a fractional picture of the complex molecular events involved, therefore limiting our understanding of the effective mechanisms mediating the effect of the exposome.

Taking examples from real-life exposome projects we will illustrate the use of statistical and machine learning techniques to accommodate co-occurring exposures contributing to population stratification, explore the links between these and cardiometabolic outcomes, and investigate the (multi)-omic response to these sets of exposures.