Statistical approaches to study exposome-health associations: from methods to pediatric health

Short bio: Augusto Anguita-Ruiz is a Junior Group Leader at the Barcelona Institute for Global Health (ISGlobal). He obtained his PhD in 2021 from the University of Granada, working on multi-omics integration and machine learning for molecular markers of insulin resistance in children with obesity. His current research line develops statistical and machine learning methods for high-dimensional, correlated and longitudinal exposome data, and applies them to pediatric cardiometabolic health within the HELIX and ATHLETE consortia. He works on outcome-wide exposome modelling, multi-omics data integration, and explainable AI, and leads the development of open methodological toolboxes, tutorials and training resources for the exposome community. Abstract: Exposome research has produced complex datasets: hundreds of correlated exposures, repeated measurements across life stages, heterogeneous data layers and multiple interrelated health outcomes. This talk presents a methodological programme addressing these challenges, and the resources it has produced for the community. I will first present a simulation study benchmarking regression and dimensionality reduction methods for repeated exposure data, which yields practical guidelines on when each approach is appropriate. I will then move from single-outcome to outcome-wide analysis, introducing mixed-response sparse reduced-rank regression (msRRR) as a method that accommodates continuous and binary repeated outcomes, high dimensionality and missing data, together with simulation work evaluating its performance. I will show how this outcome-wide exposome pipeline is applied to the HELIX cohort, identifying exposures with effects across multiple health domains. I will close with current work on exposomics and multi-omics data integration, and on the data resources and ongoing projects that could serve as a basis for collaboration with the department.