Applying AI models to actual drug R&D projects will help researchers improve the efficiency of preclinical drug discovery. The AI-driven drug discovery platform aims to cover the entire process of preclinical new drug development, including protein structure prediction, virtual screening, molecular design/optimization, ADMET prediction, and synthetic route planning.

The importance of algorithm in AI
AI helps drug research and development, and the three elements of AI—algorithm, computing power, and data, are indispensable and complementary. Advanced algorithms can deeply mine the existing big data and analyze the hidden relationships. This process not only directly assists the discovery of new drugs, but also integrates a large number of existing databases, while promoting the generation and accumulation of new data, and better optimization of algorithms. The optimized algorithm in turn can reduce the model's dependence on the amount of data and improve the model's normalization.

The role of AI in the process of new drug discovery

The process of pre-clinical new drug discovery involves the discovery and verification of targets, the discovery and optimization of lead compounds, and the identification and development of clinical candidate compounds. An AI-driven drug discovery platform usually covers the entire process of pre-clinical new drug discovery.

The first step drug target discovery and target identification and confirmation, namely finding the site of action of the drug in the body. Determining the structure of the target protein is the key task, which is regarded as an important cornerstone of drug development. For example, a protein is involved in a certain disease and becomes an important part of a key pathway. After researchers understand the structure of the protein, they can design drug molecules to regulate the function of the protein. The experimental determination of protein structure is often difficult, time-consuming, and expensive, however, with the deep learning model, the structure and function of the protein can be predicted, and by using computers, the potential lead compounds can be quickly and specifically found from hundreds of millions of small molecules.

Thanks to the breakthroughs in two key technologies, the protein structure prediction can be achieved more efficiently. One is the protein folding method based on self-supervised learning, which does not rely on homologous sequences, but directly learns the co-evolutionary model through self-supervised learning from the sequence database, so as to generate pseudo-coevolutionary information from scratch; the second is to effectively integrate template modeling and free modeling through an iterable method based on deep learning, and for the first time a dynamic and iterable amino acid pair-specific Constraints significantly improve the accuracy of modeling, and thus better fold proteins.

Screening for the target compound is the second step in new drug discovery. Compared with traditional experimental screening, virtual screening by computational methods does not need to consume compound samples, which can greatly save manpower and material resources. The ligand-based drug design method (LBDD) is one of the common methods of virtual screening. It refers to learning and establishing the relationship between molecular structure and activity from the structure of known active ligand small molecules. Because the measured compound activity data of many targets is very limited, the accuracy of the prediction model is severely restricted. AI is expected to solve this problem: for example, the virtual screening module of AI&Medicine uses meta-learning and deep neural network algorithms for the LBDD task, and "transfers" knowledge learned from other targets to the target to improve the prediction accuracy of the model. At present, the median prediction accuracy of the algorithm on thousands of experimental data sets has increased.

In the late stage of drug development, it is particularly important to predict the ADMET properties of the molecule (including drug absorption, distribution, metabolism, excretion and toxicity). According to statistics, the proportion of late drug failure caused by the nature of ADMET is as high as 60%. Therefore, early detection and elimination of molecules with poor druggability can greatly reduce the risk of late drug development failure. The AI-based ADMET property prediction allows medicinal chemists to quickly modify the molecular structure, optimize the physical and chemical properties of the molecule, shorten the drug development cycle, and reduce the cost of experimental testing.

Author's Bio: 

AI & Medicine has proudly developed a unique artificial intelligence drug research and development platform to offer drug development solutions for worldwide customers. Through big data analysis and other technical means, its AI platform can quickly and accurately mine and select the appropriate compounds or organisms.