Indicators for the Analysis of Dispersion of Datasets with Batched and Ordered Samples.

R package dispersionIndicators

Convex hull plot, convex and ICM dispersion indicators

1. Description

This package provides function for analysing the dispersion of values from a batched and ordered dataset.

It can generate convex hull plots for each batch of a given dataset, along with computing convex and ICM (Integrated Covariance Mahalanobis) dispersion indicators.

The convex hull of a set of multidimensional points is the smallest convex polygon that encloses all its points.

The convex indicators are based on convex hulls while ICM dispersion indicators are rather based on mahalanobis distance; All these provide insights into the distribution of the data points and their dispersion among and between batches.

These indicators where developped in the context of metabolomics data during the thesis of Elfried Salanon [1] that ended in 2025 (soon to be published). The convex indicators are extensively described in [2] while the ICM indicators are detailed in [3]. The main functions in this package are:

• compute_icm_distances(...): Computes ICM dispersion indicators.
• convex_analysis_of_variables(...): Performs a comprehensive analysis by computing convex hull and all convex indicators for each variable (columns) in the dataset.
• plot_all_convex_hulls(...): Generates convex hull plots for each batch in the dataset.

2. Prerequisites

R (version 4.1 or higher) is required to run this package.

The package requires the following R packages to be installed:

install.packages("ggplot2")
install.packages("stats")
install.packages("corpcor")
install.packages("utils")

3. Installation

You can install the package from CRAN using the following command:

install.packages("dispersionIndicators")

4. Usage

Load the package:

library(dispersionIndicators)

Prepare your data:

data <- data.frame(
batch = rep(c("A","B","C"), each = 10),
injectionOrder = rep(1:30, times = 1),
metabolite1 = rnorm(30, mean = 100, sd = 10),
metabolite2 = rnorm(30, mean = 200, sd = 20)
)

Run the analysis:

result <- convex_analysis_of_variables(
  data = data,
  variable_columns = c("metabolite1", "metabolite2"),
  batch_col = "batch",
  sample_order_col = "injectionOrder",
  impute_if_needed = "median",
  mode = "global"
)

Generate and save the convex hull plots:

plot_all_convex_hulls(
  target_file_path = "../convex_hulls.pdf",
  convex_analysis_res = result,
  show_points = TRUE,
  mode = "global"
)

Compute ICM dispersion indicators:

result <- compute_icm_distances(
  data,
  batch_col = "batch",
  mode = "all",
  center_method_individual = "batch",
  center_method_inter = "mean"
)

References

[1]Salanon, Elfried M.B. (2025). Reproducibility and multi-source integration in metabolomics for the identification of common phenotypes.

[2]Salanon, Elfried M.B. & al. (2024). An alternative for the robust assessment of the repeatability and reproducibility of analytical measurements using bivariate dispersion. Chemometrics and Intelligent Laboratory Systems, 250, 105148.

[3]Salanon, Elfried M.B. & al. (2025). A benchmarking workflow for assessing the reliability of batch correction methods.