New Technology to Promote Drug Development-AI Technology

The use of computer simulation for drug development may reduce costs and increase the efficiency and success rate of the drug development process. Since the 1990s, computer methods such as homology modeling, molecular docking, quantitative structure-activity relationship and molecular dynamics simulation have been used for drug development. But thanks to the rapid growth of computing power and the rise of large data sets, artificial intelligence (AI) technology has begun to be applied to enhance the process of drug development. The use of AI to develop new drugs is to simulate the process of drug screening by computer to predict the activity of compounds, and then conduct targeted physical screening of potential compounds.
Machine learning (ML)
The method of using AI in the discovery of small drug-like molecules involves the use of chemical space. By enumerating possible organic molecules by calculation, the chemical space provides a stage for the identification of brand-new high-quality molecules. Machine learning is a type of technology in artificial intelligence, which includes supervised learning, unsupervised learning and reinforcement learning. Machine learning technology helps to identify target-specific virtual molecules and their associations with their respective targets, while optimizing safety and efficacy attributes. QSAR/QSPR models are currently widely used. The applications of machine learning in drug development include but are not limited to the following: biological activity or physicochemical prediction, prediction of drug-protein and drug-drug pair interactions, and de novo molecular design with ideal pharmacological properties The molecular structure, synthesis accessibility prediction, prediction of the product of the synthesis reaction.
Deep learning (DL)
Deep learning is one of the latest technologies in the field of machine learning, which is based on the development of artificial neural networks in machine learning. This technology has shown superior performance to other machine learning algorithms in fields such as image and speech recognition, natural language processing, etc. Deep learning can process a large amount of data that has not been deeply processed manually, and automatically extract features through continuous learning for characterization simulation and classification.
Deep learning architecture for small molecule drug discovery
There are different types of DL architectures, each of which can recognize patterns and extract high-level features in different ways based on the structure of the training data.
Deep Neural Network (DNN)
DNN is a learning method with more than two hidden layers. It is used not only to process labeled data in supervised learning, but also to analyze unlabeled data in unsupervised learning. Deep Autoencoder Network (DEAN) is one of the most common generative network architectures in unsupervised learning. Recently, Generative Adversarial Networks (GAN), another DL algorithm for unsupervised learning, has been developed and widely used for image synthesis, image-to-image conversion, and super-resolution.
Convolutional Neural Network (CNN)
CNN is one of the most representative architectures in DL and is widely used in many fields such as image and speech recognition and natural language processing (NLP).
Recurrent Neural Network (RNN)
RNN is another representative architecture in DL. RNN is widely used specifically for processing sequence data. It is more suitable for some time-dependent tasks such as speech recognition in natural language processing, handwriting recognition, stock prediction and other fields.
Automatic encoder (AE)
AE is a neural network used for unsupervised learning, which aims to reduce the non-linear dimensionality, and has been widely used to learn to generate models from data.
Machine learning has proven to be a useful tool in the field of drug discovery. Compared with other methods, deep learning has a more flexible architecture, so it can be dedicated to solving a certain type of problem. These computational tools have shown outstanding performance in target identification, compound screening, de novo drug design and clinical prediction. With the advancement of artificial intelligence technology, the impact of AI on drug development will only increase.