Predict of the Hidden Drug Interactions Based on Similarity Utilizing the Knowledge Graph

Drug-drug interactions (DDIs) can significantly impact treatment outcomes by causing adverse effects or reducing therapeutic efficacy. Whereas many DDIs are already known, there are numerous hidden interactions. Therefore, this study tries to predict unknown interactions based on drug structure and target information from the Drug Bank dataset. This paper trains the random forest model integrated with a knowledge graph on a dataset containing 922 unique drugs and 731,196 drug interactions. To illustrate the complex interactions between these drugs, we created a knowledge graph with Neo4j. This graph integrates information gathered from the dataset and extracts relevant data using Cypher queries. The inference engine was created according to the model's predictions on unknown DDIs. The random forest model, with an F1-score of 0.82, achieved a recall of 0.99 and a precision of 0.70. Also, 34,580 potential DDIs were extracted from the sample dataset of drugs. By embedding the knowledge network, our inference engine significantly enhanced predicting the hidden interactions and was found strong in the balance of precision and recall. Our research is applicable to the enhancement of drug safety and effectiveness via mitigating potential DDIs by identifying them.