SURMOUNT-5 proteomics paper

Proteomics analysis code for the SURMOUNT-5 manuscript: dual-platform (Olink Explore HT + SomaScan 11K) characterization of tirzepatide vs semaglutide in patients with obesity.

Study

• Trial: SURMOUNT-5, Phase 3b, open-label, active-comparator
• Arms: Tirzepatide 15 mg or maximum tolerated dose (MTD) vs Semaglutide 2.4 mg or MTD
• Population: Adults with obesity/overweight without type 2 diabetes (N=751 randomized)
• Proteomics timepoints: Weeks 0, 24, 72
• Platforms: Olink Explore HT (5,416 assays, NPX log2) and SomaScan 11K (10,771 aptamers, RFU linear)

Repository structure

analysis/            Manuscript analysis scripts and pipeline modules
  manifest.yaml      Script/input/output manifest for paper reproduction
  pipelines/         Manuscript-facing pipeline entrypoints and study-specific glue
  scripts/           Analysis CLIs that write derived outputs
  tools/             Manifest, aliasing, and helper utilities
  qc/                Olink/Soma QC package
  mmrm/              R mmrm + emmeans wrapper
  pathway/           cameraPR + ORA wrapper
  mediation/         R-bootstrap mediation wrapper
data/                Data inventory and protected-data schema notes
figures/             Publication figure scripts
tables/              Supplementary table builders
results/             Derived MMRM/pathway/mediation outputs when locally available

Setup

This repo uses pixi.toml for the runtime toolchain and conda-managed R packages, and pyproject.toml + uv.lock for Python package resolution.

pixi install --locked
pixi run sync
pixi run check-r

pixi run sync enforces the committed uv.lock, so Python dependencies are not silently re-resolved during reproducibility checks.

Key Python dependencies: polars, pandas, somadata, ultraplot, pypdf, rpy2, scipy, scikit-learn, statsmodels.

pixi.toml pins Python 3.13.13 and R 4.5.3, then installs the R analysis packages directly from conda-forge and bioconda for the analysis.mmrm, analysis.mediation, and analysis.pathway entrypoints:

• mmrm 0.3.17
• emmeans 2.0.3
• mediation 4.5.1
• limma 3.66.0
• gdsfmt 1.46.0
• SNPRelate 1.44.0

Usage

All uv run commands below assume you are inside pixi shell. Running uv run outside of pixi shell will recreate a .venv instead of using the pixi-managed environment (which provides R, rpy2, and the locked conda packages). Start a session with:

pixi shell

Analysis scripts are run as Python modules. The public repo is CLI-first; no notebooks are required for manuscript reproduction.

# List manuscript reproduction commands
uv run python analysis/tools/publication_manifest.py list

# Validate manifest paths and static figure style conventions
uv run python analysis/tools/publication_manifest.py check
uv run python analysis/tools/publication_manifest.py audit-figures
uv run python -m compileall analysis figures tables

# The static checks above do not require mounted protected inputs.
# The analysis, figure, and table build commands below do.

# Analysis CLIs
uv run python analysis/pipelines/run_qc.py --help
uv run python analysis/scripts/build_uniprot_map.py
uv run python -m analysis.mmrm --help
uv run python -m analysis.pathway --help
uv run python -m analysis.mediation --help

# Generate main figures
uv run python figures/fig1_trajectory.py
uv run python figures/fig2_volcano.py
uv run python figures/fig3_pathway.py
uv run python figures/fig4_mediation.py
uv run python figures/fig5_ppy.py

# Build a combined main-figure PDF (Figure 1-5 labels only)
uv run python figures/build_main_figures.py

# Build the combined Supplementary Figures PDF (A4 pages, standard 0.5 in page margins, 9/8 pt legends)
uv run python figures/build_supplementary_figures.py

# Build supplementary tables
uv run python tables/build_all.py

Data Access

Clinical trial data is accessed via DNAnexus (dxfuse mount at /mnt/project/). Copy paths.example.yaml to paths.yaml and fill in the mounted roots (mmrm_root, qc_root, pheno_root, geno_root, qa_root, raw_proteomics_root). paths.py resolves surmount5_* study directories under those roots. See analysis/manifest.yaml for the manuscript script-to-output manifest.

Protected raw and participant-level data are not included. The public codebase contains the code required to regenerate manuscript-derived outputs when the protected inputs are mounted.

Citation

If you use this code, please cite the associated publication:

Chiou J, Dunn JP, Dimitriadis GK, Falcon B, Yan D, O'Dushlaine C, Coghlan M, Harris C, Titchenell P. Differential Plasma Proteomic Responses between Tirzepatide and Semaglutide in Adults with Obesity: An Exploratory Analysis of SURMOUNT-5. TODO: journal. 2026. DOI: TODO

See CITATION.cff for machine-readable citation metadata.

License

This project is licensed under the MIT License.

Copyright © 2026 Eli Lilly and Company.