Pittion et al. (2025) introduce an R package called hdmax2, which implements an enhanced version of the “High-Dimensional Mediation Analysis using the Max-Squared” (HDMAX2) method originally proposed by Jumentier et al. (2023) for high-dimensional mediation analysis. The goal of mediation analysis is to quantify the indirect effect of a variable M in the causal relationship between exposure X and outcome Y. The fundamental concept behind HDMAX2 methods is to use a latent factor mixed model to estimate the effects of unobserved confounders and a max-squared test to identify significant mediators. The HDMAX2 method represents a significant advancement in the case of high-dimensional mediation, such as DNA methylation or gene expression analysis, where the number of mediators often far exceeds the sample size.

The main contributions of this article are the implementation of the HDMAX2 method as an R package, and an extension of the original method to binary outcomes and to binary, categorical, and multivariate exposures, as opposed to only continuous variables. The package includes visualization tools, helper functions for mediator selection, and options for handling multivariate exposures. A key strength of the package lies in its versatility. The new package, hdmax2, accommodates a variety of data types. This flexibility makes it a valuable tool for researchers analyzing high-throughput molecular data. Finally to illustrate this flexibility, the authors present two case studies that were not described in the Jumentier et al. (2023) analysis. In the first case study, the authors employed mediation analysis to assess the potential causal role of DNA methylation in the pathway linking the HER2 status of breast cancer (a marker for an aggressive breast cancer subtype) to a survival risk score, which was derived from a six-gene expression signature and is inversely correlated with patient survival. In the second case study, the authors conducted mediation analysis to explore the role of gene expression in the pathway linking patient gender to the occurrence of multiple sclerosis specific subtypes: clinically isolated syndrome and relapsing-remitting multiple sclerosis. These illustrate the relevance of hdmax2 to study the transcriptome and the methylome.

In conclusion, the hdmax2 R package will be invaluable for handling high-dimensional molecular data in the study of the intricate pathways through which exposures influence health outcomes.

References

Jumentier B, Barrot C-C, Estavoyer M, Tost J, Heude B, François O, Lepeule J (2023) High-dimensional mediation analysis: A new method applied to maternal smoking, placental DNA methylation, and birth outcomes. Environmental Health Perspectives, 131, 047011. https://doi.org/10.1289/EHP11559

Pittion F, Jumentier B, Nakamura A, Lepeule J, Francois O, Richard M (2025) hdmax2, an R package to perform high dimension mediation analysis. HAL, ver. 4 peer-reviewed and recommended by PCI Genomics. https://hal.science/hal-04658960