subnetDR

subnetDR is an R/Python workflow for subtype-specific network module identification and drug repositioning. It integrates proteomics data, protein-protein interaction networks, drug response prediction, drug-target interaction prediction, and perturbation-response scoring to prioritize candidate drugs for cancer subtypes.

Software Copyright

A Subtype-Specific Network Module Identification and Drug Repositioning System Software V1.0
Software Copyright, Registration No. 2025SR142352, 2025.

Overview

Cancer heterogeneity can lead to distinct molecular subtypes with different treatment responses and clinical outcomes. subnetDR is designed to identify subtype-specific protein network modules and prioritize candidate drugs by combining:

• Proteomics-based subtype analysis
• Differential protein expression analysis
• Protein-protein interaction network construction
• Network module detection and functional annotation
• Drug response prediction
• Drug-target interaction prediction
• Elastic network model-based perturbation response analysis
• Drug repositioning score calculation and ROC evaluation

The complete workflow can be run through run_subnetDR_pipeline() or executed step by step.

Workflow

The subnetDR pipeline contains three major stages.

1. Subtype-Specific Module Identification

Scripts:

• DEPs.R
• NetConstruct.R
• ModuleDivision.R
• ModuleSelection.R
• ModuleAnnotation.R

Main tasks:

• Identify differentially expressed proteins across subtypes
• Construct subtype-specific PPI networks using String, PhysicalPPIN, and ChengF networks
• Detect network modules using Louvain or ne-PCA/WF-based module division
• Select modules based on minimum node number
• Perform GO, KEGG pathway, and Hallmark enrichment analyses

2. Drug Response Network Construction

Scripts:

• DrugResponse.R
• SEQCre.R

Main tasks:

• Predict drug response using oncoPredict
• Cluster samples based on module-specific expression profiles
• Construct drug response networks
• Extract protein FASTA sequences and drug SMILES files for downstream drug-target prediction

3. PRS-Based Drug Repositioning

Scripts:

• BindingScore.R
• PScore.R
• DPIROC.R

Main tasks:

• Predict drug-target binding affinity using DeepPurpose
• Compute perturbation response matrices using elastic network model-based analysis
• Integrate binding affinity and perturbation sensitivity into a PS score
• Rank candidate drug-target pairs
• Evaluate predictions using known drug-target interactions from TTD and DrugBank

System Requirements

The software was tested under:

• OS: Windows 10 64-bit
• CPU: Intel Core i5-7500 @ 3.40 GHz
• Memory: 32 GB RAM
• R: 4.1.1
• Python: 3.10.14

Python-related steps require a Conda environment. The example environment below uses Python 3.9 and is compatible with the package workflow described in the software manual.

Installation

Install R Package

install.packages("devtools")
library(devtools)
install_github("LilyYNY/subnetDR")
library(subnetDR)

Create Python Environment

Open Anaconda Prompt or a terminal with Conda enabled.

conda create -n py3.9 python=3.9
conda activate py3.9
pip install rdkit
pip install descriptastorus
pip install DeepPurpose
pip install seaborn
pip install goatools
pip install prody

If rdkit cannot be installed through pip, install it through Conda: conda install -c conda-forge rdkit In R, the Python environment can be selected through reticulate or passed to subnetDR functions through the py_env argument.

Required Input Files

For the automated workflow, place all required files under one project folder, for example F:/sample_test.

expression.xlsx
subtype.xlsx
GDSC2_Expr.RData
GDSC2_Res.RData
DTI_DrugBank.RData
DTI_TTD.RData

Required PPI directory:

PPI/
├── String/
│   ├── 9606.protein.info.v11.5_new.txt
│   └── 9606.protein.info.v11.5.txt
├── PhysicalPPIN/
│   └── physicalppi
└── ChengF/
    ├── 9606.protein.info.v11.5_new.txt
    ├── Homo_sapiens_gene_info
    └── Human Interactome

Required ENM directory:

python/
└── enm/

The input phenotype file subtype.xlsx should contain at least:

Sample
Subtype

The expression matrix expression.xlsx should contain genes or proteins in the first column and samples in the remaining columns.

Quick Start

library(devtools)
install_github("LilyYNY/subnetDR")
library(subnetDR)

base_path <- "F:/sample_test"
py_env <- "C:/Users/YangMiao/.conda/envs/py3.9"

run_subnetDR_pipeline(
  base_path = base_path,
  py_env = py_env
)

The pipeline runs the full workflow from differential expression analysis to network construction, module detection, functional annotation, drug response prediction, binding affinity prediction, perturbation-response scoring, and ROC evaluation.

Notes

Please prepare PPI databases, GDSC drug response data, DrugBank/TTD drug-target data, and ENM files before running the full pipeline. See the example sample_test workflow for expected file organization. The package can be run as a full automated pipeline or step by step for debugging and customization.