CFDE Lung Transcriptomics Course

This is a five-module course in lung transcriptomics built on Common Fund Data Ecosystem (CFDE) resources. You will discover datasets through CFDE surfaces, download them from DCC portals, and carry one analysis from raw nuclei through quality control, cross-donor integration, differential expression and pathway enrichment, paired multiome interpretation, and a cross-ecosystem contrast comparison.

Primary omics files are not included and are not redistributed. The folders they belong in are already named and empty; you download into them. Module 0 sets up the environment and Module 1 covers discovery, download and verification, with terminal commands for every file in its Appendix A.

Attribution

Copyright (c) 2026 Brian Billings.

Title. CFDE lung transcriptomics course (HuBMAP / GTEx bridge with GEO and OSDR arms).

Licenses. Course prose, documents, notebooks, and derived tables and figures: CC BY 4.0. Code under scripts/: MIT. Third-party gene-set libraries under data/genesets/ remain under their upstream terms; see THIRD-PARTY-NOTICES.md.

What is in this package

Included	Not included
Six module documents, under pdfs/	HuBMAP / GTEx / GEO / OSDR primary files
Notebooks 01 to 04, outputs cleared	Alternate HuBMAP blocks and SNARE products other than Donor_2 HBM828.GPVG.252
Analysis package under scripts/ and config/paths.yaml	Regenerated data/processed/*.h5ad
Empty named folders for every teaching HuBMAP ID	Full optional MuData (~14.7 GiB)
Committed figures and reports under outputs/	Tables the pipeline regenerates when it runs
Gene sets, ortholog and ERCC reference files

outputs/tables/ holds only the files the pipeline reads rather than writes: the two Azimuth label caches, the committed GeneLab differential-expression subset, the NES sensitivity table, the offline discovery caches, and the download manifest. Everything else appears when you run the notebooks.

Documents

Module	Document	Time	Notebook
0	pdfs/module0.pdf	20 min reading	none, environment setup and code orientation
1	pdfs/module1.pdf	45 min	notebooks/01_discovery_and_qc.ipynb
2	pdfs/module2.pdf	45 min	notebooks/02_integration_de_pathways.ipynb
3	pdfs/module3.pdf	30 min	notebooks/03_multiome.ipynb
4	pdfs/module4.pdf	30 min	notebooks/04_cross_ecosystem.ipynb
5	pdfs//module5.pdf	30 min	none, communication and assessment

Three hours of instruction across Modules 1 to 5, excluding Module 0.

What you need

Requirement	Value
Python	3.11
Environment	conda/mamba CFDE_lung_env, or a pip venv from requirements.txt
RAM	16 GB recommended; 8 GB minimum (measured peaks: Module 2 DE 7.6 GB, Module 3 7.1 GB, Module 4 5.5 GB)
Disk	5 GB, covering the download plus the objects the notebooks write
Download	About 815 MB across 20 files, itemized in Module 0 Appendix B with retrieval commands in Module 1 Appendix A
Gene sets	Shipped under data/genesets/ (Hallmark, Reactome, GO BP; no KEGG)

Quick start

mamba env create -f environment.yml
mamba activate CFDE_lung_env
python -m ipykernel install --user --name cfde_lung_env --display-name "Python (CFDE_lung_env)"

Read docs/module0.docx first: it covers the environment, the code layout, and the ten Python patterns the analysis packages use. Then work Module 1, which walks discovery, download and verification before any analysis runs.

python scripts/verify_access.py          # confirm every taught URL is reachable
jupyter lab notebooks/01_discovery_and_qc.ipynb

Pip alternative:

python3.11 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

Cohort

Donor_1, Donor_2, Donor_3, Donor_4, four HuBMAP lung donors selected for comparative value across age, sex and reported race and ethnicity.

Donor	Age	Sex	snRNA block	Also provides
Donor_1	37	Male	HBM347.VDHS.379	Module 1 QC scaffold
Donor_2	25	Female	HBM473.NKMR.872	SNARE-seq2 multiome HBM828.GPVG.252 (Module 3)
Donor_3	56.8	Female	HBM484.RGWP.797
Donor_4	52.96	Male	HBM379.QCJT.435

Module 4 adds GTEx v10 lung, GEO series GSE150910, and NASA OSDR study OSD-248.

Interpreting without running

Notebooks ship without stored cell outputs on purpose, so a clean re-run is forced after download. The results are readable before you run anything:

1. docs/module{0-5}.docx, the module documents, with figures and tables embedded
2. outputs/figures/, every figure the documents cite
3. outputs/reports/, methods records and biological interpretation drafts

Design notes

• No primary data redistribution: DCC portals for files, CFDE surfaces for discovery.
• GTEx enters at Module 4, not Module 2.
• Gene sets: Hallmark, Reactome and GO BP only (no KEGG).
• Module 4's headline is mechanism concordance against a permutation null; the global NES correlation is supporting context only.
• Every number in the documents is read from a committed table rather than transcribed.

Optional smoke run

After the download is complete:

python scripts/run_modules_smoke.py --module all