Methods for Single and Multilayer Network Analysis for Biomedical Research: The mixmashnet R Package

Abstract

INTRODUCTION
Network medicine provides a framework through which many biomedical phenomena are currently understood, reflecting the multidimensional nature of available data (e.g., genes, diseases, functional parameters). This is particularly relevant in aging research, where aging is driven by interconnected biological processes spanning multiple physiological domains. However, most existing network analysis tools are limited to single layer or homogeneous data structures, and uncertainty quantification is rarely addressed.
OBJECTIVE
Develop a statistical framework for the estimation and interpretation of heterogeneous single and multilayer networks and to demonstrate its application in aging-related biomedical data.
METHODS
We developed an R package, MixMashNet, for the estimation and analysis of heterogeneous single and multilayer networks using Mixed Graphical Models (MGMs). The MGM framework enables the joint modelling of continuous, categorical, and binary variables, allowing the definition of distinct yet interconnected layers. Importantly, the estimation procedure incorporates regularization penalties specifically designed to address the challenges of high-dimensional data. MixMashNet supports both intra and interlayer connections, offering control over which layers can interact and which remain independent, while allowing adjustment for selected covariates.
The package provides tools for community detection using different algorithms, estimation of centrality and bridge-centrality indices, and computation of community scores. In addition, non-parametric bootstrap procedures are implemented to evaluate uncertainty of community assignments, edge weights and node centralities. MixMashNet also enables the identification of excluded nodes that may act as bridging nodes between biological processes. Visualization tools based on ggplot2 and Shiny are included for interactive network exploration. 
RESULTS
The proposed framework enables the estimation and characterization of complex dependency structures in biomedical data while incorporating uncertainty quantification through bootstrap procedures. MixMashNet supports the identification of stable network organization, community structure, and node-level properties, facilitating the interpretation of multidimensional biological systems. In particular, the framework enables the identification of bridge nodes connecting different biological domains, including nodes excluded from stable community assignment that may still contribute to network connectivity. MixMashNet was applied to several biomedical datasets, including aging-related biomarker data from the Swedish National Study on Aging and Care in Kungsholmen (SNAC-K) cohort, illustrating its potential for investigating interactions across different domains such as chronic diseases and blood-based biomarkers. 
CONCLUSION
MixMashNet provides a unified framework for the estimation, visualization, and interpretation of heterogeneous single and multilayer networks with integrated uncertainty quantification. By supporting the investigation of complex biological mechanisms and their connections, MixMashNet may facilitate the generation of clinically interpretable hypotheses in network medicine, aging research and precision medicine.