How bio-ontology-research-group/dti-voodoo works
DTI-Voodoo is a research pipeline for predicting candidate drugs for a protein by combining drug molecular features, ontology-derived phenotypic effects, protein interaction networks, and graph neural networks. The principal workflow prepares interaction and feature data, builds model inputs, predicts drug–target interactions, and evaluates predictions. The README documents the combined method and its capabilities; sampled source supports data preparation, ontology graph embedding, and protein-function prediction, but does not expose the complete runtime wiring, so uncertain call relationships are omitted.
Main entry points: torch_dti_predictor.py Graph DTI predictor · data Biological datasets · DTI_data_preparation.py DTI preparation · DL2vec Ontology embeddings
Components 19
Data and preparation
- Biological datasets
data - DTI preparation
src/DTI_data_preparation.py
Feature generation
- PhenomeNET features
src/PhenomeNET_DL2vec_utils.py - Ontology embeddings
src/DL2vec - Ontology graph builder
src/DL2vec/generate_graph.py - Graph walk embeddings
src/DL2vec/compute_vectors.py - Molecular features
src/molecular_utils.py - SMILES models
src/smiles_transformer - Protein function prediction
src/protein_function_predictor.py - Protein representation
models/protein_representation - PPI network utilities
src/PPI_utils.py
DTI prediction
- Graph DTI predictor
src/torch_dti_predictor.py - Neural network layers
src/torch_networks.py - DTI model utilities
src/torch_dti_utils.py - Molecular DTI predictor
src/molecular_predictor.py - Candidate interactions
results
Evaluation
- Prediction evaluation
src/results_evaluation.py - DTI metrics
src/dti_utils.py
Other components
- Researcher
Connections 22
- Researcher → DTI preparation: initiates
- Biological datasets → DTI preparation: supplies data
- DTI preparation → Graph DTI predictor: prepares inputs
- DTI preparation → Molecular DTI predictor: prepares inputs
- Biological datasets → Ontology embeddings: supplies ontology data
- PhenomeNET features → Ontology embeddings: uses embeddings
- Ontology embeddings → Ontology graph builder: builds graph
- Ontology embeddings → Graph walk embeddings: generates vectors
- Biological datasets → Molecular features: supplies drug data
- SMILES models → Molecular features: encodes SMILES
- Biological datasets → Protein function prediction: supplies sequences
- Protein representation → Protein function prediction: provides representation
- Biological datasets → PPI network utilities: supplies interactions
- PhenomeNET features → Graph DTI predictor: provides phenotype features
- Molecular features → Graph DTI predictor: provides molecular features
- PPI network utilities → Graph DTI predictor: provides network data
- Graph DTI predictor → Neural network layers: uses layers
- Graph DTI predictor → DTI model utilities: uses utilities
- Graph DTI predictor → Candidate interactions: writes predictions
- Molecular DTI predictor → Candidate interactions: writes predictions
- Candidate interactions → Prediction evaluation: evaluates
- Prediction evaluation → DTI metrics: uses metrics