pguIMP is an R package that provides a comprehensive, interactive pipeline for reproducible preprocessing and cleaning of bioanalytical laboratory data. Developed in collaboration between data scientists and bioanalytical diagnostics experts, pguIMP enables field experts and researchers to prepare high-dimensional bioanalytical data for machine learning and advanced statistical analysis without requiring programming expertise.
The package implements a fixed sequence of preprocessing steps within an intuitive Shiny-based graphical user interface, ensuring reproducible and traceable data preprocessing workflows.
Data preprocessing is critical for the success of machine learning and statistical analyses of biomedical data. The evaluation of bioanalytical data by classical statistics, machine learning approaches, or pharmacokinetic-pharmacodynamic modeling depends heavily on data quality. However, most available preprocessing tools either:
pguIMP solves these problems by offering:
✓ Interactive Graphical Interface – No programming required
✓ Machine Learning Methods – Advanced imputation and outlier detection
✓ Visual Validation – Real-time visualization of preprocessing results
✓ Reproducibility – Complete documentation of all preprocessing decisions
✓ Flexibility – Choose from multiple algorithms for each preprocessing step
✓ Traceability – Full audit trail with seed values and session reports
install.packages("pguIMP")
# Install remotes package if needed
if (!requireNamespace("remotes", quietly = TRUE))
install.packages("remotes")
# Install development version
# remotes::install_github("JornLotsch/pguIMP")
install.packages("pguIMP", repos = c('https://jornlotsch.r-universe.dev', 'https://cloud.r-project.org'))
library(pguIMP)
# Launch the Shiny-based web interface
pguIMP::pguIMP()
This opens an interactive graphical interface where you can:
library(pguIMP)
# Load your data
data <- read.csv("your_data.csv")
# Create a data object
pgu_data <- pgu.data(data)
# Transform variables (example: log transformation)
transformed <- pgu.transformator(pgu_data, method = "log")
# Impute missing values using kNN
imputed <- pgu.imputation(transformed, method = "knn")
# Normalize the data (example: z-score)
normalized <- pgu.normalizer(imputed, method = "zscore")
# Export results
pgu.exporter(normalized, file = "preprocessed_data.csv")
| Function | Purpose |
|---|---|
pguIMP() |
Launch interactive Shiny application |
pgu.data() |
Create and initialize data object |
pgu.explorer() |
Explore data distributions and structure |
pgu.transformator() |
Apply data transformations |
pgu.normalizer() |
Normalize/scale data |
pgu.outliers() |
Detect outliers using various methods |
pgu.imputation() |
Impute missing values |
pgu.limitsOfQuantification() |
Handle values below LOQ separately |
pgu.validator() |
Validate preprocessing results |
pgu.exporter() |
Export preprocessed data and reports |
pgu.reporter() |
Generate comprehensive preprocessing reports |
pguIMP has been rigorously evaluated on real bioanalytical datasets from:
Results demonstrate that:
✓ Machine learning-based imputation methods (kNN, Random Forests) significantly outperform simple mean/median substitution, especially for Not-At-Random missing data
✓ Proper preprocessing improves downstream clustering accuracy and biomarker identification
✓ The pguIMP pipeline preserves data structure better than alternative approaches
✓ Integration into machine learning workflows enhances model performance
If you use pguIMP in your research, please cite the following publication:
Malkusch, S., Hahnefeld, L., Gurke, R., & Lötsch, J. (2021). Visually guided preprocessing of bioanalytical laboratory data using an interactive R notebook (pguIMP). Clinical and Translational Science, 14(6), 2213-2224. https://doi.org/10.1111/cts.13089
Full Paper: PDF available on PubMed Central
@article{Malkusch2021,
title={Visually guided preprocessing of bioanalytical laboratory data using an interactive {R} notebook ({pguIMP})},
author={Malkusch, Sebastian and Hahnefeld, Lisa and Gurke, Robert and L{\"o}tsch, J{\"o}rn},
journal={Clinical and Translational Science},
volume={14},
number={6},
pages={2213--2224},
year={2021},
doi={10.1111/cts.13089}
}
help(pguIMP) after installation?function_name()
Core Development Team:
Contributing Authors:
For questions or issues, please contact the maintainer
This project is licensed under the GNU General Public License v3.0 or later (GPL ≥ 3).
See the LICENSE file for details.
This work was funded by the Landesoffensive zur Entwicklung wissenschaftlich-ökonomischer Exzellenz (LOEWE), specifically the LOEWE-Zentrum für Translationale Medizin und Pharmakologie (J.L.), through the project "Reproducible Cleaning of Biomedical Laboratory Data Using Methods of Visualization, Error Correction and Transformation Implemented as Interactive R-Notebooks."
The funders had no role in method design, data selection and analysis, decision to publish, or preparation of the manuscript.
pguIMP was developed in close collaboration between:
The package implements contemporary data processing and machine learning methods while maintaining a focus on usability for domain experts without programming expertise.
Key references for methods implemented in pguIMP:
Last Updated: February 2026
Package Version: Check DESCRIPTION file for current version