How sujithreddy-47/oncograph works
OncoGraph is a TCGA pan-cancer analysis framework: it preprocesses gene-expression and clinical data, builds a patient-similarity graph, trains and evaluates a graph neural network, then supports subtype, biomarker, survival, and anomaly analyses. The README documents these capabilities and the overall methodology; sampled entry points and the source index clarify several dispatch relationships. Some downstream implementations and data-loading details were not sampled, so the graph omits uncertain internal links.
Main entry points: data_processor.py Preprocessing · train_model.py Train Command · evaluate_model.py Evaluate Command · graph_builder.py Similarity Graph
Components 22
Data and Graph
- TCGA Inputs
EB++AdjustPANCAN_IlluminaHiSeq_RNASeqV2.geneExp.xena - Clinical Inputs
Survival_SupplementalTable_S1_20171025_xena_sp - Preprocessing
backend/utils/data_processor.py - Similarity Graph
backend/utils/graph_builder.py - Processed Data
backend/data/processed_data.pt - Patient Graph
backend/data/graph_data.pt
Model Lifecycle
- Model Factory
backend/models/model_factory.py - GCN Architectures
backend/models/gcn_models.py - Training Loop
backend/training/trainer.py - Train Command
backend/train_model.py - Model Evaluation
backend/evaluation/evaluator.py - Evaluate Command
backend/evaluate_model.py
Clinical Analysis
- Subtype Discovery
backend/analysis/subtypes.py - Biomarker Discovery
backend/analysis/biomarkers.py - Survival Analysis
backend/analysis/survival.py - Anomaly Detection
backend/analysis/anomaly_detection.py - Subtype Command
backend/run_subtypes.py - Biomarker Command
backend/run_biomarkers.py - Survival Command
backend/run_survival.py - Anomaly Command
backend/run_anomaly.py
Configuration
- Runtime Config
backend/configs/config.py
Other components
- Researcher
Connections 26
- Researcher → Train Command: starts training
README.md - TCGA Inputs → Preprocessing: expression input
README.md - Clinical Inputs → Preprocessing: clinical input
README.md - Preprocessing → Processed Data: saves processed
backend/utils/data_processor.py - Processed Data → Similarity Graph: builds from
backend/utils/graph_builder.py - Similarity Graph → Patient Graph: saves graph
README.md - Train Command → Runtime Config: loads config
backend/train_model.py - Train Command → Similarity Graph: loads graph
backend/train_model.py - Train Command → Model Factory: creates model
backend/train_model.py - Train Command → Training Loop: starts training
backend/train_model.py - Model Factory → GCN Architectures: instantiates
backend/models/model_factory.py - Researcher → Evaluate Command: starts evaluation
README.md - Evaluate Command → Model Evaluation: runs evaluation
backend/evaluate_model.py - Evaluate Command → Model Factory: creates model
backend/evaluate_model.py - Researcher → Subtype Command: runs analysis
README.md - Subtype Command → Subtype Discovery: dispatches to
backend/run_subtypes.py - Researcher → Biomarker Command: runs analysis
README.md - Biomarker Command → Biomarker Discovery: dispatches to
backend/run_biomarkers.py - Researcher → Survival Command: runs analysis
README.md - Survival Command → Survival Analysis: dispatches to
backend/run_survival.py - Researcher → Anomaly Command: runs analysis
README.md - Anomaly Command → Anomaly Detection: dispatches to
backend/run_anomaly.py - Subtype Discovery → Preprocessing: uses data
backend/analysis/subtypes.py - Biomarker Discovery → Preprocessing: uses data
backend/analysis/biomarkers.py - Survival Analysis → Preprocessing: uses data
backend/analysis/survival.py - Anomaly Detection → Preprocessing: uses data
backend/analysis/anomaly_detection.py