This algorithm performs batch effect and cohort effect correction on intensity data using a statistical modeling approach based on linear mixed models (LMM). This tool is especially suited for omics or high-throughput datasets, where batch or injection order can introduce unwanted variability.

The correction includes:

• Log transformation of intensities
• Batch-wise standardization (z-score normalization)
• Linear mixed-effects modeling using injection order and random batch effects
• Residual extraction and back-transformation of the signal

✨ Features:

• Handles multiple intensity columns simultaneously
• Corrects for both injection order and batch-specific effects
• Optional visualization (if ggplot2 and patchwork are enabled)
• Integrates with command-line tools and Galaxy workflows

If desired, before/after boxplots of batch effects can be visualized using ggplot2. These are currently commented out in the code but can be easily activated.

Install from: https://cran.r-project.org/

install.packages(c("dplyr", "lme4", "ggplot2", "patchwork", "optparse"))

• OS: Windows, MacOS, or Linux
• R ≥ 4.0 recommended

To run the algorithm from terminal:

Rscript ./parsec.R -i data_test.csv -o data_output.csv

• -i or --input: CSV file with columns Batch, SampleID, Injection_Order, and intensity columns (e.g., Ion1, Ion2).
• -o or --output: Name for the processed output CSV file.

Batch,SampleID,Injection_Order,Ion1,Ion2
A,1,1,120,230
A,2,2,150,260
B,3,1,110,200
B,4,2,140,240

mutate(across(all_of(intensity_cols), log1p))

Improves normality of data distribution.

group_by(batch) %>% mutate(across(intensity_cols, ~ scale(.x)[, 1]))

Ensures mean = 0 and SD = 1 within each batch.

Ion ~ Injection_Order + (1|Batch)

Accounts for systematic cohort effects while removing batch bias.

mutate(across(all_of(intensity_cols), expm1))

Returns corrected intensities to original scale.

• Elfried Salanon : [email protected]
• Marie Lefebvre : [email protected]

@softwareversion{swh-dir-f166076,
    author = "Salanon, Elfried  and Comte, Blandine  and Boccard, Julien  and Pujos-Guillot, Estelle ",
    organization = "Université Clermont Auvergne, INRAE, UNH, Plateforme d’Exploration du Métabolisme, MetaboHUB Clermont, Clermont-Ferrand, France and School of Pharmaceutical Sciences, University of Geneva, Geneva, Switzerland",
    license = "CC-BY-NC-4.0",
    abstract = "This algorithm performs batch effect and cohort effect correction on intensity data using a statistical modeling approach based on linear mixed models (LMM). This tool is especially suited for omics or high-throughput datasets, where batch or injection order can introduce unwanted variability.",
    date = "2025-06-13",
    year = "2025",
    month = jun,
    file = "https://github.com/inrae/PARSEC",
    repository = "https://github.com/inrae/PARSEC",
    title = "PARSEC",
    version = "1.0.0",
    swhid = "swh:1:dir:f16607641d45c7ae90571e847660d9040ca5723f;origin=https://github.com/inrae/PARSEC;visit=swh:1:snp:6ead07484c3aa11bdecba951b42a5d037e1cbf0c;anchor=swh:1:rev:51c7d9c1e08354ffcccb208e3f355a4cd6a9d60b"
}