Title: Non-linear Mediation Analysis with High-dimensional Mediators whose Causal Structure is Unknown

Abstract: With multiple potential mediators on the causal pathway from a treatment to an outcome, we consider the problem of decomposing the effects along multiple possible causal path(s) through each distinct mediator. Under Pearl's path-specific effects framework (Pearl, 2001; Avin et al., 2005), such fine-grained decompositions necessitate stringent assumptions, such as correctly specifying the causal structure among the mediators, and there being no unobserved confounding among the mediators. In contrast, interventional direct and indirect effects for multiple mediators (Vansteelandt and Daniel, 2017) can be identified under much weaker conditions, while providing scientifically relevant causal interpretations. Nonetheless, current estimation approaches require (correctly) specifying a model for the joint mediator distribution, which can be difficult when there is a high-dimensional set of possibly continuous and non-continuous mediators. In this article, we avoid the need for modeling this distribution, by building on a definition of interventional effects previously suggested by VanderWeele and Tchetgen Tchetgen (2017) for longitudinal mediation. We propose a novel estimation procedure that uses non-parametric estimates of the marginal (counterfactual) mediator distributions. The procedure is illustrated using publicly-available genomic data exploring the prognostic effect of microRNA miR-223 expression on the three-month mortality of patients suffering from an aggressive form of brain cancer, that is potentially mediated by expression values of multiple genes (Huang and Pan, 2016).