How does Life science move towards AI4S era?

New methods and new tools have opened up a new chapter in the field of life Science, and now the arrival of AI for Science is bringing unprecedented new opportunities to the field of life science. At the academic summit of 2023 Science Intelligence Summit "Al for Life Sciences" held on August 11, excellent experts and scholars at home and abroad shared their insights on key scientific issues of AI for Life Sciences. This paper discusses how life science moves towards AI4S era.

AI has become an important factor driving drug development

Zhang Ao, distinguished professor and dean of School of Pharmacy, Shanghai Jiao Tong University, believes that with the rapid development of life science technology and disease biology, the biomedical research paradigm and industrial development pattern are undergoing complex and profound changes. In particular, with the rapid evolution of advanced computing technology and artificial intelligence technology, the use of massive data and advanced methods to accelerate drug development and promote the evolution of computing resources to the natural stage has become a development trend, and the innovation of AI computing tools has become an important factor driving drug research and development.

Drug discovery is a multi-stage, lengthy and expensive process, and efficiency gains at each stage have enormous value. With the blessing of data-driven, "AI+ drug research and development" is becoming an inevitable development trend.

Professor Yang Shengyong of the State Key Laboratory of Biotherapeutics, West China Hospital, Sichuan University, has been engaged in computer-aided drug molecular design research for a long time. In recent years, he has carried out in-depth research in AI-based drug molecular design, especially in molecular generation models. For example, the team established a novel molecular generation model based on conditional recurrent neural networks and applied this model to the development of RIPK1 small molecule inhibitors, discovering a highly active and highly selective RIPK1 inhibitor with a novel structure, which demonstrated the ability of generative deep learning (GDL) models to generate novel molecular structures. It shows that deep learning has great potential in the field of drug discovery. The study was published in Nat Commun (2022).

"Our team introduced artificial intelligence algorithms into drug molecular design and established a new method of drug molecular design based on capsule neural network, using vector neurons instead of traditional scalar neurons, which can solve the small sample problem encountered in drug research and development to a certain extent, and the learning effect of data is better. Significantly improve the accuracy of drug molecular design and pharmacokinetic prediction. We have also recently developed molecular generation models based on conditional recurrent neural networks and autoregressive flows, which have demonstrated a good ability to generate novel structurally drug-like molecules." Yang Shengyong said.

AI is significantly empowering the overall drug discovery process. By using AI technologies such as deep learning and reinforcement learning, researchers can quickly screen potential drug molecules, predict drug targets, and optimize drug molecular design, greatly improving the efficiency and success rate of drug research and development.

How to make AI better serve drug development? Ouyang Defang, a professor at the Chinese Institute of Medicine, University of Macau, has been focusing on computational pharmacy research since 2011, establishing a drug preparation database and developing a machine learning algorithm for drug preparation prediction, thereby building an artificial intelligence platform for drug preparation, and combining quantum mechanics, molecular simulation and pharmacokinetic simulation for drug dosage form research.

Ouyang Defang said that modern pharmacy from "physical pharmacy", "nanomedicine" to "computational pharmacy", the field of pharmacy has made great progress in the past decade, but there are still many challenges. The first is the lack of high-quality data, the current pharmaceutical factory is still a data island, how to share data is very important. The second is that there is still a lack of user-friendly digital tools or computing tools for prescription developers, and there is an urgent need to develop user-friendly computing tools. The third is interdisciplinary personnel training, high-quality personnel is the fundamental to promote the development of this field.

The rapid development of machine learning based on data modeling and innovative intelligent computing tools will continue to meet the increasingly diverse analysis needs and promote the transformation of the life sciences industry. In general, the paradigm of drug discovery is undergoing tremendous disruption and change.

AI4S new paradigm injects new vitality into life science research

Artificial intelligence is playing an increasingly important role in the study of biological mechanisms, the screening, detection and treatment of diseases. With the completion of the Human Genome Project, scientists have a better understanding of the relationship between disease and genes.

In the field of genomics, through machine learning and statistical models, researchers have successfully predicted important biological processes such as the function of encoded proteins in the genome, gene shear events, and regulatory networks, providing a huge data resource for biological science research.

Under the new paradigm of AI4S, the pre-trained model shows unprecedented capabilities, but for specific scientific data and information, it also needs to deeply combine the underlying characteristics of the discipline, and use its special "language" as a carrier, such as protein sequence and nucleic acid sequence in life science, which is different from natural language to carry information.

In RNA drug design, researchers urgently need a computational tool that can efficiently and comprehensively explore and describe the RNA space in order to achieve digital innovation in RNA research. Uni-RNA developed by Deep potential technology came into being. According to Wen Han, head of macromolecular research and development of Deep Potential Technology, Uni-RNA uses about 1 billion high-quality RNA sequences for large-scale pre-training, covering almost all RNA space, and fully mining the potential information of RNA sequences. By fine-tuning the model in a wide range of downstream tasks, Uni-RNA achieved leading results (SOTA) in all seven major tasks in the three RNA domains: RNA structure prediction, mRNA sequence property prediction, and RNA function prediction, opening up endless possibilities for future in-depth research in the RNA field.

"Al for Life Sciences", a new paradigm that combines AI capabilities with underlying biological mechanisms, is injecting new vitality into the entire industry from the perspective of underlying technological breakthroughs, and its systematic development is expected to bring more possibilities to the industry.

Finding neoantigens is the core of personalized cancer immunotherapy. However, using traditional experimental methods to find precise recognition of T-cell-bound Peptides on the cell surface and then validate the "immunogenicity" is still a costly and time-consuming way. Li Ming, a fellow of the Royal Society of Canada and a professor at the University of Waterloo in Canada, said, "The specificity of neoantigens between individuals is extremely high, and it is difficult to get an accurate match in the general protein or peptide database, so de novo sequencing technology without a database reference requires a high degree of accuracy, and an efficient method is urgently needed to verify the neoantigens." The introduction of deep learning methods and the development of "neoantigen de novo sequencing" methods have brought new opportunities. Simulating the human central tolerance system to solve the problem of no TCR in immunogenicity prediction, using deep learning technology, the detection accuracy and efficiency of neoantigens are greatly improved. In the field of life sciences, artificial intelligence helps promote the combination of "wet experiments" and "dry experiments" to promote the landing and universal benefit of personalized cancer immunotherapy."

The use of AI technology for data analysis, model construction and prediction, as well as the development of computational tools, all demonstrate the great potential and broad application space of AI4S in the field of life sciences.

Zhang Dan, associate professor at Central China Normal University, said that "iron death" is a programmed way of cell death that is different from apoptosis, necrosis and autophagy. Catalase (CAT) is expected to be a new target to induce iron death in tumor cells, and the key to achieve this goal is to design and synthesize highly active CAT inhibitors. The results of animal experiments showed that BT-Br could effectively inhibit tumor growth in CRPC mice. This study suggests that CAT has the potential to be a new target for the treatment of CRPC based on iron death induction strategies. This set of processes can be applied to the field of plants for weeding or fighting fungal diseases, which is also the result of the application of artificial intelligence technology to the laboratory from calculation to synthesis, to molecular biology experiments and animal and plant experiments.

Today, the new paradigm of AI for Science has provided strong support for the field of life science and greatly promoted the innovation and development of the field of life science. With the continuous progress of AI technology, future life science research will be more intelligent and efficient, bringing more hope for human health and well-being.

On August 10-11, the 2023 Science Intelligence Summit was successfully held in Beijing. As a series of activities of the Zhongguancun Forum, the 2023 Science Intelligence Summit is hosted by the Beijing Institute of Science Intelligence, aiming to build a co-creation platform for scientific research breakthroughs, technology cultivation and talent exchange in the field of AI for Science. The summit set up a main forum and 10 thematic academic summits, covering topics such as model algorithms, databases, energy materials, and computing engines. At the meeting, participating academicians, experts and business representatives shared advanced ideas and cutting-edge insights, presented research results and innovative technologies, and looked forward to the future development trend of AI for Science.