Drug-Similarity-and-Link-Prediction-using-Graph-Embeddings-on-Medical-Knowledge-Graph
Utilizing graphical neural networks and embeddings on a medical database KEGG to perform link predictions and drug similarity system.
Section of KEGG dataset visualized as knowledge graph (using NetworkX)
Different color represent different entity present in the database that are Drugs, Diseases, Gene, Network and Pathway:
![alt text](https://github.com/prakhargurawa/Drug-Similarity-and-Link-Prediction-using-Graph-Embeddings-on-Medical-Knowledge-Graph/raw/main/images/KG.png?raw=true)
Graph Neural Network pipeline to generate graphical embeddings, link prediction and drug similarity
![alt text](https://github.com/prakhargurawa/Drug-Similarity-and-Link-Prediction-using-Graph-Embeddings-on-Medical-Knowledge-Graph/raw/main/images/pipeline.png?raw=true)
Graph embedding visualized in 2D space using PCA
![alt text](https://github.com/prakhargurawa/Drug-Similarity-and-Link-Prediction-using-Graph-Embeddings-on-Medical-Knowledge-Graph/raw/main/images/ComplEx2DEntity.png?raw=true)
Comparision of different graphical embedding models on link prediction tasks
![alt text](https://github.com/prakhargurawa/Drug-Similarity-and-Link-Prediction-using-Graph-Embeddings-on-Medical-Knowledge-Graph/raw/main/images/algo-compare.png?raw=true)