ProteinBatcher

ProteinBatcher is a Bioconductor-style R package for standardized, design downstream analysis of label-free quantitative proteomics tables (with a focus on DIA-NN-style DIA outputs). It provides utilities to read, validate and harmonize quantitative proteomics tables, convert them to SummarizedExperiment, run imputation-aware differential abundance testing with limma, and optionally generate publication-ready volcano plots and deregulograms.

---

Why ProteinBatcher?

Bioconductor already provides excellent tools for statistical modeling and container infrastructure (notably limma and SummarizedExperiment), but end-to-end proteomics workflows that combine:

• missingness-based filtering,
• mean/LDV imputation tailored to label-free proteomics,
• flexible multi-variable designs (batch/sex/other covariates),
• correlation-aware models via duplicateCorrelation,
• and standardized output objects for downstream reporting

are less common in a single, reproducible pipeline.

ProteinBatcher focuses on orchestrating these steps in a consistent way, so repeated analyses across experiments become less error-prone and more comparable.

Core features

1) Standardized pipeline wrapper

The main entry point is:

• run_proteomics_pipeline()

It performs: import -> missingness filtering -> imputation -> limma testing -> output organization -> (optional) plots and tables.

2) Imputation-aware differential testing

test_limma_customized() fits limma models and stores results in rowData(se).

Key feature: BH-adjusted p-values are recomputed after excluding features imputed as LDV in both groups of a contrast (LDV/LDV). This avoids artificially significant results driven by left-censored imputations.

3) Flexible experimental designs (including interactions)

You provide a formula (e.g. ~ 0 + condition + batch + condition:batch) plus contrasts (tests) and optional interaction tests (tests_interaction).

4) Paired and correlation-aware designs

• paired = TRUE: ensures replicate is included in the design for paired/technical repeated measures.
• block_effect = TRUE: uses limma::duplicateCorrelation() with colData(se)$block and fits correlation-aware models.

5) Plotting for interpretation

• plot_volcano_customized() produces volcano plots from SummarizedExperiment results stored in rowData().
• plot_deregulogram() produces deregulograms for 2-level interaction factors (e.g. female/male, batch 1/batch 2), comparing full effects between levels.

Important interpretation note: the deregulogram is not expected to label the exact same proteins as the interaction volcano plot. The interaction volcano highlights significant interaction coefficients; the deregulogram emphasizes a more interpretable subset where interaction occurs in the context of statistically significant condition/full effects.

---

Installation

Bioconductor (recommended once available)

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

BiocManager::install("ProteinBatcher")
``