ECM GRADES

Overview

ECM GRADES is a multi-omic analysis pipeline for ECM-based patient stratification and network-driven drug prioritization in lung adenocarcinoma (LUAD). Using proteogenomic data from the CPTAC cohort, the pipeline constructs ECM-guided patient-specific networks, derives ECM barcodes for patient clustering, and performs network proximity analysis to identify candidate therapeutic targets.

The repository includes all input data, intermediate outputs, and analysis code needed to reproduce the results reported in the associated manuscript.

---

Repository Structure

ecm_grades/
├── data/
│   ├── DRUGBANK_14_03_2024/
│   ├── cbioportal/luad_cptac_2020/
│   ├── clinical-cptac-3.2023-10-02/
│   ├── lists/
│   ├── raw_data/
│   └── reference_interactome/
├── network_modeling/
│   ├── patients/
│   ├── reference/
│   ├── scripts/
│   ├── terminals/
│   ├── final_parameter_list.csv
│   ├── ecm_grades.sh
│   └── ecm_momix.yml
├── network_proximity/
│   ├── input_data/
│   ├── network_proximity_pool.py
│   ├── proximity_results_insol.csv
│   └── proximity.yml
├── out_data/
│   ├── TRRUST/
│   ├── enrichments/
│   ├── lists/
│   ├── network_data/
│   ├── pptx_data/
│   ├── terminals/
│   └── trx_data/
├── patients_all/
├── consensus_networks.py
└── main.rmd

Directory / File	Description
data/	Raw and preprocessed input data
data/DRUGBANK_14_03_2024/	FDA-approved drug data, drug targets, and annotations for network proximity analysis
data/cbioportal/luad_cptac_2020/	Omic and clinical data from CBioPortal
data/clinical-cptac-3.2023-10-02/	Patient survival and recurrence data from CPTAC
data/lists/	Gene and drug lists — Matrisome genes, cancer-associated pathway genes, OncoTreat drug list
data/raw_data/	Raw omic data from Gillette et al. (2020) supplementary materials
data/reference_interactome/	Raw and processed reference PPI interactome for network analyses
network_modeling/	Scripts for building ECM-guided patient-specific networks
network_modeling/patients/	Per-patient directories with OmicsIntegrator2 solutions across all parameter combinations
network_modeling/scripts/	Auxiliary scripts: get_memberships.py, get_parameterlist.R, get_networks.py
network_modeling/terminals/	Prize files for each patient
network_modeling/ecm_grades.sh	Bash pipeline script orchestrating the auxiliary scripts
network_modeling/ecm_momix.yml	Conda environment for network modeling
network_proximity/	Network proximity analysis inputs and code
network_proximity/input_data/	Reference interactome, consensus networks, and drug-target data
network_proximity/network_proximity_pool.py	Runs the network proximity analysis
network_proximity/proximity_results_insol.csv	Relative proximity values between ECM consensus networks and drug targets
network_proximity/proximity.yml	Conda environment for network proximity analysis
out_data/	All generated outputs — networks, tables, and figures
out_data/TRRUST/	TF enrichment results per patient from TRRUST
out_data/enrichments/	Pathway enrichment results (KEGG, REACTOME, HALLMARK, GO Biological Processes)
out_data/lists/	Fold-change values, normalized expression data, and ECM barcodes (multi-omic and single-omic)
out_data/network_data/	Network statistics — average cost, edge/node/terminal counts
out_data/pptx_data/	Differentially phosphorylated ECM proteins per patient (used as terminals alongside TFs)
out_data/terminals/	Final terminal lists per patient
out_data/trx_data/	Differentially expressed ECM genes per patient for TF enrichment
patients_all/	GraphML files of ECM-guided patient-specific networks
consensus_networks.py	Constructs ECM consensus networks from patient-specific networks
main.rmd	Main analysis workflow (R Markdown)

---

Requirements

R (>= 4.1)

Package	Package	Package
rmarkdown	tidyverse	readxl
ggstatsplot	ggpubr	M3C
vegan	survival	survminer
immunedeconv	enrichR	ComplexHeatmap
patchwork	circlize	reshape2
ggforestplot	ggvenn	gridExtra

Python (>= 2.7)

Package	Package
OmicsIntegrator	numpy
pandas	networkx

Exact package versions are provided in the supplementary tables of the associated manuscript. Conda environments for each pipeline stage are included as .yml files.

---

How to Run

1. Network Modeling

Activate the conda environment and set network_modeling/ as the working directory. Place patient prize files in terminals/ and the reference interactome in reference/, then run:

conda activate ecm_momix  # from ecm_momix.yml
bash ecm_grades.sh

Outputs are written to patients/.

---

2. Building Consensus Networks

From the root ecm_grades/ directory, ensure patient-specific networks are in patients_all/ as .graphml files and ECM grade annotations are under out_data/lists/. Then run:

python consensus_networks.py

Outputs are written to out_data/network_data/.

---

3. Network Proximity

Activate the conda environment and set network_proximity/ as the working directory. Place the reference PPI interactome, consensus networks, and drug-target data in input_data/, then run:

conda activate proximity  # from proximity.yml
python network_proximity_pool.py

Outputs are written to the working directory.

---

4. Downstream Analyses

Open main.rmd in RStudio. This script handles:

• Input data preparation for network modeling and proximity analyses (terminal preparation, PPI filtering, drug filtering)
• Multi-omic data analysis
• ECM barcoding and patient clustering
• Pathway and TF enrichment analyses
• Clinical and survival analyses
• Cellular deconvolution
• Generation of all manuscript figures

All outputs are saved to out_data/ with self-descriptive filenames.

---

Contact

For questions or issues, please open an issue or contact the repository maintainers via email (adansik22@ku.edu.tr or ntuncbag@ku.edu.tr).

Citation

Dansık, A., Sarıca, S., Öztürk, E., & Tuncbag, N. (2025). Extracellular matrix-guided tumor stratification and network models reveal clinical molecular grades. bioRxiv. https://doi.org/10.1101/2025.08.29.672994