Ramin amiri - facult of Mathematics, Statistics and Computer Science Tabriz university Tabriz, Iran
Zahra Baniasad - faculty of Mathematics and Computer Shahid bahonar university Kerman, Iran

Drug-drug interactions may cause irreversible drug side effects. Therefore, the importance of identifying drug-drug interactions before prescribing multiple drugs is quite clear. Clinical diagnosis of drug-drug interactions generally requires a lot of time and money. Computational methods as an alternative provide a much cheaper way to identify large-scale interactions. Most methods only predict whether one drug interacts with another. But they do not examine the extent of the interaction effects. Examining the relationship between two variables and interaction-effect is very important, because it will help to diagnose and understand the performance of a drug that has a strategic role in prescribing drugs. In this method, using structural relationships between drugs, we have considered a set of interactions of specific drugs. Then we have designed a new network with a combination of features and in-depth learning to predict the harmful effects of a pair of drugs together.Evaluation of the interaction-harmful results of AUC = ۰.۹۳ and AUP = ۰.۸۷ shows that the superiority of this method compared to the work done in the past

Harmful interaction, Effects of drugs, Deep learning, Machine learning