Olink® PEA Analysis Workflow

AnalyzeOlink provides a configurable report-based framework for analyzing Olink® Proximity Extension Assay (PEA) data, focusing on differential abundance testing and downstream validation. Supported statistical methods include Type III ANCOVA and linear mixed-effects models (via the afex package), t-tests and Wilcoxon tests (paired and unpaired), as well as Kruskal–Wallis and Friedman tests.

The package includes nine R Markdown analysis reports that cover key workflow components: assumption checks for ANOVA and LMER models, leave-one-out sensitivity analyses for model terms and sample-level robustness, ELISA validation, non-parametric covariate effects, and NPX correlations with metadata variables.

An annotations module integrates Human Protein Atlas (HPA) metadata, including tissue- and cell-type–specific expression profiles and Tau specificity scores. These annotations can be used to color interactive downstream visualizations such as volcano plots and estimate-vs-estimate plots.

Each phase of the analysis is implemented as an .Rmd report and can be coordinated by a Snakemake pipeline. The central .Rds outputs are created by:

• differential_expression.Rmd: Outputs the PEA .Rds required by all _pea.Rmd scripts.
• elisa.Rmd: Outputs the ELISA .Rds required by all _elisa.Rmd scripts.

Analysis Components

PEA-Focused Reports

Report	Description
differential_expression.Rmd	QC, contrast configurations, and differential abundance testing.
assumptions_pea.Rmd	Parametric assumption checks for the PEA models (normal errors, homoscedasticity, outliers).
sensitivity_pea.Rmd	Sensitivity of model results to specific samples and model terms (leave-one-out analysis).
covariates_pea.Rmd	Tests whether covariates (e.g., Age, Sex) differ between contrast arms (for use with non-parametric frameworks).
correlations_pea.Rmd	Spearman correlations between NPX values and selected metadata variables.
annotations_pea.Rmd	Human Protein Atlas (HPA) IHC tissue-specific annotations of PEA results.
functional_enrichment_pea.Rmd	Functional enrichment analysis (ORA / GSEA). Note: Currently under development.

ELISA-Focused Reports

Report	Description
elisa.Rmd	QC and differential analysis of ELISA measurements.
assumptions_elisa.Rmd	Parametric assumption checks for the ELISA models.

---

Quickstart using Test Data on Linux

1. Clone the GitHub repository

git clone git@github.com:NICHD-BSPC/AnalyzeOlink.git

2. Create the Conda environment

cd environment/
conda env create -p ./env --file env.yaml -y

3. Activate the environment

conda activate ./env

4. Generate test data

# From repo root
cd ../
Rscript data/test_data.R

5. Run the pipeline

cd snakemake
snakemake -j 8

6. View results

Open the .html reports generated in the rmds/ directory in a web browser. Each report links to its exported tables and figures.

---

Resource Requirements and Runtime

• Installation time: < 3 minutes to clone the repo, build the Conda environment, and create the test data.
• Workflow runtime (test data): < 3 minutes to run the Snakefile.
• Hardware requirements:
  • ≥ 2 GB RAM
  • ≥ 2 CPU cores

Tested Platforms

• Operating system: Linux (Ubuntu LTS; tested on GitHub Actions ubuntu-latest runners, as of December 2025: Ubuntu 24.04)
• Conda distribution: Miniforge3 (GitHub Actions latest Miniforge3 release, as of December 2025)
• Conda version: 25.11.1

---

User Configuration (for your project's data)

To run the workflow on your own PEA/ELISA data, you must configure the pipeline in two locations.

1. Global Settings (config/config.yaml)

This is the primary control center for the pipeline. Open config/config.yaml and read the detailed comments inside.

The config file handles:

• Input Paths: Locations for your Olink and optionally ELISA data files.
• Metadata Requirements: Specifics on required columns (e.g., SampleID, Grouping columns).
• Pipeline Control: Which analysis reports to run or skip via Snakemake.
• Global Parameters: Significance thresholds (alpha), parallelization (cores), and enrichment databases.

2. Analysis-Specific Settings (.Rmd Scripts)

While file paths are handled globally, specific analysis logic is configured directly within the individual R Markdown reports.

Open the .Rmd files in rmds/ and look for the code chunks containing the text, "USER SETTINGS".

Key configurations found in .Rmd files include:

• Outlier Removal: Defining specific sample IDs to exclude based on QC.
• Contrast Definitions: Defining the specific statistical comparisons (e.g., disease_vs_control) and subsetting the data for each contrast.
• Covariate Selection: Specifying which metadata columns to use for correlation or covariate testing.
• Sensitivity Checks: Defining which model terms to test in leave-one-out analyses.

---

Output Summary

Paths defined under output_paths in config/config.yaml are populated with:

• XLSX tables
• PDF figures

The .html reports in rmds/ provide clickable links to these outputs.

---

Dependencies and Package Versions

All package dependencies are managed via Conda in environment/env.yaml. Use the same environment to run both Snakemake and the .Rmd scripts (if running them outside of Snakemake). See the Conda specification file at environment/env.yaml for exact package versions and dependency resolution.

---

Support

For questions or issues, please reach out to the NICHD Bioinformatics and Scientific Programming Core.

---

License

This project is licensed under the GNU General Public License v3.0 (GPL-3.0). See the LICENSE file for full terms.