nutriNetwork

nutriNetwork implements copula graphical models (CGMs) in R to learn direct multivariate associations in nutritional epidemiology datasets and reconstruct conditional-independence networks linking nutrient intakes and health outcomes.

Key Features:

• Copula Graphical Models (CGMs): Implements CGMs to model complex associations in multivariate dietary datasets and infer direct associations beyond marginal correlations.
• Precision Matrix Inference: Infers a precision matrix whose nonzero elements indicate conditional dependencies among variables, representing direct associations.
• Data type accommodation: Handles continuous, discrete, and mixed data types within the same modeling framework.
• Application to Dutch older-adult cohorts: Reconstructs networks from Nu-Age, ProMuscle, ProMO, and V-Fit data and identifies partial correlations between vegetable protein, folate, vitamins B-6 and B-12 and physical functioning measures such as the Short Physical Performance Battery (SPPB) and chair rise test.

Scientific Applications:

• Uncover internutrient relationships: Reveals direct associations among nutrients consumed together, improving interpretation of dietary patterns in nutritional epidemiology.
• Link nutrients to physical functioning: Identifies nutrient associations with functional outcomes in older adults, including SPPB and chair rise test scores, to inform hypotheses on diet–health interactions.

Methodology:

Uses copula graphical models to learn the structure of direct associations, infers a precision matrix to represent conditional independence, and corrects for the effects of remaining variables to uncover pairwise associations while accommodating continuous, discrete, and mixed data types.