A team of researchers from the Gwangju Institute of Science and Technology (GIST) in South Korea have developed a deep learning model that predicts drug-drug interactions (DDIs) based on their effects on gene expression. DDIs can be a serious problem when multiple drugs are taken at the same time, leading to adverse health effects due to unexpected interactions.
The research was published in the Journal of Cheminformatics.
Early Detection of DDIs
Many complex diseases require the prescription of multiple drugs, or polypharmacy. With that said, the ingestion of multiple drugs can lead to all sorts of unexpected and undesired interactions, which can result in severe side effects or decreased clinical efficacy. In order to prevent patients from encountering such adverse effects, these DDIs must be detected early.
Current approaches involve computational models and neural network-based algorithms that examine prior records of known drug interactions before identifying the structures and side effects they are associated with. However, these systems assume that similar drugs have similar interactions and identify drug combinations with similar adverse effects.
The team set out to develop a new model to get around some of these limitations. The team was led by associate professor Hojung Nam and Ph.D. candidate Eunyoung Kim from GIST. They developed a deep-learning model to predict DDIs based on drug-induced gene expression signatures.
The model, named DeSIDE-DDI, consists of two parts:
- First Part: A feature generation model that predicts a drug’s effect on gene expression. It does this by considering both the structure and properties of the drug.
- Second Part: A DDI prediction model that predicts various side effects that result from drug combinations.
“Our model considers the effects of drugs on genes by utilizing gene expression data, providing an explanation for why a certain pair of drugs cause DDIs,” Prof. Nam says. “It can predict DDIs for currently approved drugs as well as for novel compounds. This way, the threats of polypharmacy can be resolved before new drugs are made available to the public.”
All compounds do not have drug-treated gene expression signatures, so the new model relies on a pre-trained compound generation model to generate expected drug-treated gene expressions.
“This model can discern potentially dangerous drug pairs, acting as a drug safety monitoring system. It can help researchers define the correct usage of the drug in the drug development phase,” Prof. Nam continues.
The new model is a huge step forward in improving the safety of novel drugs, and it will provide much needed insight into DDIs and their adverse effects.