Statistical approaches for quantifying the mediating role of omics markers between the exposome and health
Pysyvä osoite
Verkkojulkaisu
DOI
Tiivistelmä
With increasing availability of sequencing methods, a wide range of omics markers have redeemed a central role in health sciences. The exposome, i.e. the range of life course infuences from external environment, life style and internal environment, are known to affect health and also infuence various omics markers. These markers in turn can play a role in health and disease, and thus they hold promise as mediating links between the exposome and health outcomes.
Causal mediation analysis provides a statistical framework for quantifying indirect effects of exposures carried through mediators, i.e. variables that serve as mechanistic pathways transmitting the effects of exposures on health outcomes. Some properties of omics data pose challenges in the conduct of mediation analysis. For example, sequencing count data may not be suitable in their raw form but require appropriate methods to account for their compositional nature. Omics data also often exhibit sparsity (excess zero counts). Furthermore, in high-dimensional omics datasets, the meaningful signals need to be identifed from a large number of variables to quantify the mediating role of multiple simultaneous mediators.
This thesis is motivated by empirical research questions concerning the role of the epigenome and gut microbiome in the association between exposome and health, both within individuals and across generations. The objective is to develop statistical approaches in the frameworks of mediation analysis and compositional data analy sis for investigating the mediating role of various omics markers in the relationship between the exposome and health. To this end, a method for compositional media tion analysis is presented and its performance is evaluated using extensive simulation studies and empirical data. An asymptotic normal approximation for compositional log-ratio coordinates is derived and its applicability under varying levels of sparsity is investigated in a simulation study. Compositional methods are found to perform well when sparsity is not extreme. In the empirical analyses, the methods presented in this thesis are applied to examine the mediating role of the gut microbiome and DNA methylation between exposures and cardio-metabolic health of an individual. Finally, the potential application of these methods in investigating questions concerning paternal infuences via sperm epigenome across generations is discussed.
Sarja
Turun yliopiston julkaisuja - Annales Universitatis Turkuensis, Ser. AI: Astronomica, Chemica, Physica, Mathematica|755
Saavutettavuusominaisuudet
Navigointi mahdollista, kuvilla vaihtoehtoiset kuvaukset, taulukot saavutettavia, looginen lukemisjärjestys