FibroTarget-Liver

FibroTarget-Liver is a reproducible single-cell workflow for human liver fibrosis, MASH, and cirrhosis target discovery. It starts from public count matrices, runs a Seurat-based analysis, validates priority candidates in orthogonal public data, enriches targets with public evidence, and packages the results as an executive report, ranked tables, figures, a Shiny dashboard, and a local/AWS Nextflow demo.

Primary discovery uses GSE136103, the Ramachandran et al. human cirrhosis scRNA-seq dataset. Validation uses GSE244832 for MASH/HSC biology, GSE207310 for bulk NAFLD/NASH directionality, and the excluded GSE136103 blood and mouse libraries for specificity and preclinical conservation checks.

Read This First

1. Executive summary: the short report, key results, translational interpretation, and next steps.
2. Analysis walkthrough: technical methods, why each choice was made, what was inferred, and where the caveats are.
3. Biology primer: quick grounding for the liver cell compartments and why they matter in fibrosis.
4. Written responses: answers to the eight technical questions.
5. Interactive dashboard: browser-hosted Shiny app for UMAP, candidate table, scoring, DE, validation, and QC views. Local run notes are in dashboard/README.md.
6. Nextflow demo: local and AWS-ready reproducibility demo.

Supporting details are consolidated in docs/technical_appendix.md.

What This Answers

The analysis covers the requested outcomes:

Outcome	Where to look
Dataset and metadata curation	data/metadata/gse136103_sample_manifest.csv, executive summary
QC and preprocessing choices	workflow/03_compact_analysis.R, analysis walkthrough
Major liver cell-type annotation	marker dot plot, refined labels, analysis walkthrough
Fibrosis/cirrhosis-associated genes and states	pseudobulk DE tables, executive summary
Pathway or mechanism analysis	Hallmark pathway table plus pathfindR Reactome active-subnetwork results from pseudobulk DE
Rule-based biomarker prioritization	scoring method and ranked candidate tables
Ranked 10-20 candidates	translational ranked candidate table and dashboard
Translational relevance	executive summary, written responses, evidence-enriched tables
Reproducibility	Makefile, renv.lock, Dockerfile, Nextflow demo

Main Results

The strongest fibrosis signal is a scar niche made of:

• activated stromal and HSC/myofibroblast-like cells
• scar-associated endothelial programs
• macrophage injury and repair states

The top candidates are best read by use case:

Use case	Candidates	Interpretation
Fibrosis burden and pharmacodynamic readout	COL1A1, COL3A1, TIMP1	Strong scar biology; collagens are markers, not direct targets
Secreted biomarker	SMOC2, TIMP1	Practical assay potential; TIMP1 needs specificity checks
Scar vascular niche	ACKR1, PLVAP	Strong tissue-state markers; targeting needs vascular safety work
Therapeutic hypothesis	PDGFRA, PDGFRB	Druggable stromal receptor biology; safety window is central
Macrophage validation queue	TREM2, CD9, SPP1, GPNMB	Important disease-state biology; needs macrophage atlas and spatial validation

The ranking intentionally separates biomarker value from therapeutic target value. A gene can be a strong fibrosis marker and still be a poor intervention point.

Run Locally

Requirements:

• R 4.6.0
• Seurat 5.5.0
• renv
• Java and Nextflow for the demo workflow
• internet access for public GEO downloads if raw data are not already present

Run the full local workflow:

make all

Run the dashboard:

Rscript -e "shiny::runApp('dashboard')"

Run the standalone Nextflow demo:

make nextflow-demo

Validate the repo structure:

make validate-repo

Key Files

workflow/03_compact_analysis.R          Seurat preprocessing, clustering, UMAP, compartment calls
workflow/07_refine_annotations.R        Published-reference label refinement
workflow/08_pseudobulk_de.R             Donor-level pseudobulk DE
workflow/04_prioritize_targets.R        Candidate scoring and pathway enrichment
workflow/13_validate_blood_mouse_markers.R
                                        Blood specificity and mouse ortholog validation
reports/executive_submission_summary.Rmd
                                        Source for rendered executive summary
reports/tables/ranked_biomarker_target_candidates_translational.csv
                                        Final ranked candidate table
dashboard/app.R                         Shiny dashboard
nextflow/fibrotarget_demo/              Local/AWS demo pipeline

Data Policy

Tracked:

• code and configuration
• metadata manifests
• demo data
• compact result tables and figures
• dashboard-ready CSVs
• reports and documentation

Not tracked:

• raw GEO archives
• extracted validation matrices
• large Seurat objects
• logs
• private notes

References

• Ramachandran et al. Resolving the fibrotic niche of human liver cirrhosis at single-cell level. Nature, 2019.
• Rinella et al. Multisociety Delphi consensus statement on steatotic liver disease nomenclature. Hepatology, 2023.
• GSE136103, GSE244832, and GSE207310 GEO records.