Advances on Molecular Design of Antiviral Drug Based on Deep Learning

Recently, artificial intelligence (AI) has been recognized as an effective method that has capacity to contribute to drug discovery. Deep learning (DL) is a significant category of AI. As the environment changes, the virus emerges in endless, which has confirmed the necessity of discover and design novel antiviral drug. But conventional methods of drug discovery are time-consuming, high-cost and risky. This study focuses on the principle of deep learning while using models based on several frameworks, which have been applied in severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). DL has generated some models with functions of predicting and screening efficient target or designing novel molecular. An overview of DL advancement is provided, along with suggestions for improvement and potential future directions. Although in current studies the DL models have increasingly developed, the problem of the limitation of database is still a staple hinderance. To solve this problem, introducing transfer learning, updating the target protein inhibitor database and establishing a shared database may work. Therefore, those models can achieve high accuracy and effectiveness in developing SARS-CoV-2 drugs with low-cost and low-risk. In the future, they may also be expanded to develop drugs against other viruses.