DrugNet: Network-Based Drug–Disease Prioritization by Integrating Heterogeneous Data
January 2015
in “
Artificial Intelligence in Medicine
”
Studysummary This study found that DrugNet, a network-based prioritization method, effectively improves drug repositioning tasks, achieving high performance in validation tests and clinical trial comparisons.
Our plain-language summary. Not medical advice or a treatment recommendation. Consult a qualified healthcare professional before changing treatment. Full disclaimer
The document presents DrugNet, a methodology for drug repositioning that integrates data from drug, protein, and disease networks. DrugNet was tested using cross-validation and real clinical trials, achieving high mean AUC values of 0.9552 in 5-fold cross-validation and 0.8364 for clinical trial-based tests, indicating its effectiveness in discovering new drug uses. The method outperformed another tool, PREDICT, in ranking new drug-disease hypotheses and showed promise in case studies, identifying potential new uses for existing drugs in various diseases. Despite some limitations, such as data source biases, DrugNet is seen as a valuable tool for both drug-disease and disease-drug prioritization, with the potential to save resources in drug development. Future improvements include adding new networks and similarity measures to enhance performance. DrugNet is available as a web tool, and its Matlab source code is provided.
