MetaboDynamics: a framework of probabilistic models to analyze longitudinal metabolomics data

Graphical abstract

Data requirements

• MetaboDynamics requires at least duplicates of metabolite abundances for all analyzed time points and conditions. No minimum number of metabolites is required.
• Data can be a data frame (see Vignette "using MetaboDynamics with data frames"") or a SummarizedExperiment object (demonstrated in this vignette).
• Metabolite abundances have to be stored in column named "metabolite", time points in "time" and the experimental condition in "condition"
• For clustering of metabolite abundance dynamics are at least two time points required
• To conduct over-representation analysis on KEGG functional modules KEGG IDs of metabolites are required
• To conduct comparison of metabolite abundance dynamics between clusters at least two experimental conditions are required
• If an in vitro experiment is analyzed and cell counts should be incorporated in model, cell counts have to be estimated (i.e. cell counter not counting every cell) as a poisson distribution is assumed. At least one replicate of cell counts per time point and condition is required.

Metabolite abundances have to be adequately normalized and scaled. The expected scaling is explained in the Vignette.

In the case of an in vitro experiment with available cell counts for all time points and conditions no scaling or normalization is necessary as this will be handled by the fit_dynamics_model() function (model="raw_plus_counts")

Installation

MetaboDynamics is an R package available from Bioconductor.

URL: https://www.bioconductor.org/packages/release/bioc/html/MetaboDynamics.html

To install MetaboDynamics, start R (4.5) and enter:

if (!require("BiocManager", quietly = TRUE))
    install.packages("BiocManager")

BiocManager::install("MetaboDynamics")

You can also install the development version (current bug fixes and added features can be found in the NEWS file) of MetaboDynamics:

if (!require("BiocManager", quietly = TRUE))
    install.packages("BiocManager")

BiocManager::install(version="devel")
BiocManager::install("MetaboDynamics")

Overview

MetaboDynamics facilitates the analysis of longitudinal metabolomics data e.g. from untargeted LC-MS. Common tools mostly only allow the comparison between two time points instead of analyzing the full observed dynamics profile of metabolite concentrations over multiple time points. Furthermore common tools mostly only allow to compare two experimental conditions at a single time point and are using frequentist statistical methods. As metabolomics data is often noisy, robust methods for the estimation of mean metabolite concentrations per time point are needed.

MetaboDynamics allows longitudinal analysis over multiple time points and experimental conditions employing three probabilistic models:

1. A hierarchical Bayesian model for the robust estimation of means at every time point despite varying spread between time points. Hierarchical Bayesian models are known to balance between over- and underfitting, allowing to gain as much information from the noisy data as possible while not being overly confident about the estimates. Its outputs are A) differences between time points for every metabolite (differential concentrations), B) estimated differences between metabolite abundance dynamics between different conditions, and C) metabolite specific dynamics profiles.

The Bayesian model returns posterior estimates (distributions of possible values of the estimated parameters). It's output therefore can be used to conduct clustering of metabolite abundance dynamics with bootstrapping (function cluster_dynamics()), allowing to estimate the reliability of clusters of metabolite abundance dynamics.

2. Over-representation analysis of KEGG-functional modules such as amino acid metabolism or KEGG pathways in dynamics clusters with a quantitative model that employs a hypergeometric distribution and reports probabilities of a functional module or pathway being over-represented in a cluster. Can also estimate simultaneously under-representation of functional modules.

3. Estimation of the dynamics similarity between metabolite clusters of different experimental conditions with a Bayesian model. This model infers the mean pairwise euclidean distance of composing metabolite dynamics between two clusters (i.e. every metabolite dynamics from cluster A is compared with every metabolite dynamics of cluster B). In combination with the comparison of metabolites that compose two clusters this allows to spot differences and similarities between experimental conditions. For examples clusters of metabolites with similar metabolite composition but different dynamics between experimental conditions.

With this framework the following questions can be answered:

Workflow

For a worked example on a simulated data see Vignette or if package is installed:

browseVignettes("MetaboDynamics")