NetCoupler: Inferring causal pathways between high-dimensional metabolic data and external factors

Luke W Johnston1, Clemens Wittenbecher2
1. Steno Diabetes Center Aarhus, Denmark, 2. Harvard University, USA

Luke W. Johnston, MSc, PhD
ORCID iD icon0000-0003-4169-2616
Creative Commons License
DDA Logo


Background: Metabolic data, especially from -omics, is challenging to meaningfully analyze and visualize. Even more so when we’re interested in potential causal pathways. Our aim is to develop an algorithm and R package to estimate and visualize these pathways, in order to be used by the wider research community.

Methods: The NetCoupler algorithm estimates causal pathways between a network of metabolic variables and either 1) an “outcome” variable (i.e. influenced by the network), 2) an “exposure” variable (i.e. influences the network), or 3) both. The R package is being developed at and aims to be user-friendly.

The algorithm implementation follows these steps, visually shown in Figure 1:

  1. The network skeleton of metabolic variables is constructed
  2. Each node in the network is selected
  3. All combinations of neighbours connected to the node are formed and used for adjustment in subsequent models
  4. Multi-model estimation occurs and links are classified as either no effect, ambiguous, or as direct effects based on model estimates

Example of an analysis on the UK Biobank dataset using this algorithm is shown in Figure 2.

Conclusion: We hope the NetCoupler package will give researchers more tools to meaningfully analyze complex -omic style data.


Figure 1: NetCoupler algorithm (R package at process, identifies potential pathways between exposure (E), metabolic network (N), and outcome (O)

Figure 2: Example analysis of NetCoupler identifying pathways between stature (marker of early growth), network, and HbA1c in the UK Biobank