Drug design and discovery are generally time consuming, costly, and difficult. Therefore, in recent years, the development of computational approaches to drug design and discovery has been considered. Identifying drug-target interaction is an essential step in the early stages of drug discovery, and most methods developed to date use binary classification to predict the interaction between a drug and its target. However, predicting the strength of a connection between a drug and its target is more instructive but challenging. Binding affinity indicates the strength of drug-target pair interactions.
The time and financial costs for directly measuring binding affinity through experimental methods are enormous. Therefore, computational methods have recently been considered to predict drug target binding affinity, but the use of structural information has increased the time cost and resource requirement in these methods.
In this seminar, we propose a new model to address some of the challenges of this issue. We focus on methods based on artificial intelligence (deep learning) that use the sequence of molecules as input. The results showed that the use of deep learning and transformers on target sequences improved the performance of drug-target binding affinity prediction models.