Drug Sensitivity Scoring Pipeline Based on PPI Topology and Expression Change

This repository contains the core analytical scripts used in our study titled:
"[Title of the Manuscript]", submitted to npj Precision Oncology.

It integrates transcriptomic data, protein-protein interaction (PPI) network topology, and ChEMBL-based drug-target information to prioritize therapeutic compounds in CDK4/6 inhibitor-resistant cancers.

---

🔍 Overview of Workflow

1. Differential Gene Expression
   Using either DESeq2 or limma depending on dataset format.

2. PPI Network Construction
   Filtered high-confidence STRING interactions mapped to DEGs.

3. Network Topology Analysis
   Calculating centrality scores (Degree, Betweenness, Eigenvector, MCC, EPC) and computing PCA-based weights.

4. Target Integration
   Mapping genes to UniProt, merging logFC and centrality, computing per-target importance.

5. Drug Sensitivity Scoring
   Matching with ChEMBL drug-targets and scoring drug efficacy using expression-weighted topology.

6. Statistical Validation of Scoring System
   Binomial test and graphical visualization for evaluating prediction accuracy across datasets.

---

📁 Repository Contents

File Name	Description
01_DESeq2_DifferentialExpression.R	Differential gene expression analysis using DESeq2 or limma based on counts.csv and metadata.csv
02_PPI_NetworkConstruction.R	Construct STRING-based protein-protein interaction networks
03_PPI_Topology_Analysis_and_Weighting.R	Analyze network topology and calculate centrality-based scores
04_DEG_Mapping_and_Uniprot_Merge.R	Merge DEG results with UniProt mapping and interaction partners
05_Drug_Target_Matching_and_Scoring.R	Match with ChEMBL targets and compute weighted drug scores
06_Validation_Pvalue_ScoringSystem.R	Validate prediction accuracy using binomial test and rank-based p-values (input: Summary of Datasets accuracy .csv)
counts.csv / metadata.csv	Input expression matrix and sample metadata
GSE200029_parental_vs_Resistant_filteredDEGs.xlsx	Processed differential gene expression results (example dataset)
gse20029_target_importance.csv	Target-level topology-integrated importance scores
drug_scores_gse20029.csv	Final ranked drug sensitivity scores

---

🚀 Reproducibility Instructions

# Clone the repository
git clone https://github.com/qiankey759/drug_sensitivity_scoring.git

# Run scripts in sequence
source("scripts/01_DESeq2_DifferentialExpression.R")
source("scripts/02_PPI_NetworkConstruction.R")
source("scripts/03_PPI_Topology_Analysis_and_Weighting.R")
source("scripts/04_DEG_Mapping_and_Uniprot_Merge.R")
source("scripts/05_Drug_Target_Matching_and_Scoring.R")
source("scripts/06_Validation_Pvalue_ScoringSystem.R")

All scripts are self-contained and annotated. Output files are saved in-place under results/ and figures/.

---

🔐 Data Availability

Due to the size and proprietary nature of the original datasets, only representative files are provided in this repository.
Full data and intermediate files are available upon reasonable request from the corresponding author.

---

📩 Contact

For academic correspondence or access to full datasets and drug scoring results, please contact:

Dr. Qian Keyang
Email: qiankey759@163.com
Affiliation: Wuxi Medical College of Jiangnan University