How kleberpointexe/cancerdrugrepo-ic works
This repository analyzes drug–cancer associations using protein-interaction networks and matrix factorization. Its main result paths are network proximity scoring or NMF-based scoring, followed by per-cancer rankings and evaluation against RepoDB ground truth; it also provides graph diffusion, feature-matrix construction, and plotting. The README is empty, and the excerpts do not show a top-level workflow wiring these modules together, so the graph preserves supported local calls without inferring cross-module orchestration.
Main entry points: GuneyModel.py Pair proximity · matrices.py Feature matrices · nmf.py Two-matrix NMF · metrics.py Evaluation metrics
Components 15
Data and graph
- Data loader
data/loader.py - Input datasets
data/dadosOriginais - Graph diffusion
graph/diffusion.py - Feature matrices
features/matrices.py
Network proximity
- Proximity scoring
models/GuneyModel.py - Distance precompute
models/GuneyModel.py - Pair proximity
models/GuneyModel.py
Matrix factorization
- Two-matrix NMF
models/nmf.py - Pair enumeration
features/pairDrugCancerGuney.py
Ranking and evaluation
- Drug rankings
evaluation/ranking.py - Ground truth
evaluation/ground_truth.py - Evaluation metrics
evaluation/metrics.py
Utilities and plots
- Result plots
visualization/plots.py - Experiment I/O
utils/io.py - Pickle storage
utils/pickle.py
Connections 7
- Feature matrices → Feature matrices: uses perturbation
features/matrices.py - Proximity scoring → Distance precompute: precomputes distances
models/GuneyModel.py - Proximity scoring → Pair proximity: processes pairs
models/GuneyModel.py - Pair proximity → Pair proximity: measures distance
models/GuneyModel.py - Two-matrix NMF → Two-matrix NMF: trains each k
models/nmf.py - Evaluation metrics → Evaluation metrics: scores AUC
evaluation/metrics.py - Experiment I/O → Pickle storage: uses pickle helpers
utils/io.py