As an applied statistician, my research interest lies on the interface between mathematical and epidemiological sciences. It resides on the identification, application, and implementation of statistical methods to answer specific epidemiologically driven questions. This comprises application of well-established approaches from other fields as well as the generalization and refinement of existing methods.
Specifically, I have been working over the past years on the analysis and profiling techniques from different OMIC platforms: genetic, epigenetic, transcriptomics, proteomic, and metabolomic. Method used included univariate models coupled with multiple testing correction strategies, dimensionality reduction techniques, and variable selection approaches.
My recent work focused on the integrative analysis arising form several OMIC profiles simultaneously, and explore the correlation structures within and across OMICs profiles. These approaches have the potential to contribute to the elucidation of the way the effect of external stresses is biologically mediated at the cellular level, and therefore to identify effective effect of these exposure.
Validation and refinement over the hypothesized mechanisms requires additional methodological effort, which covers a wide range of methods: from causal models, structural equation models, to Bayesian hierarchical models.
Longitudinal data (i.e. exposure history, or health trajectories) are data of choice to explore in further details the mechanisms characterizing the pathophysiological pathways leading to increased risk of an adverse condition. Their modelling relies on the analysis of trajectories using advanced time series approach, and trajectory classification models (dynamic time warping approaches), or explicit modelling of the trajectories through (multi-state) compartmental models, which are defined by a set of ordered states (compartments) reflecting the health status which can either be observed or hidden.