Figure: The AI ​​Automated Laboratory (A-Lab) system at the University of California, Berkeley, can discover and synthesize compounds on its own. \Internet image

Artificial intelligence (AI) technology is applied to drug development, which has attracted great attention in the pharmaceutical industry. Artificial intelligence not only shortens the time required for drug development, improves the success rate of research, but also saves money. With the development and growth of the AI ​​pharmaceutical industry, the first innovative drug developed by AI technology may soon be available in the near future. However, before that, AI pharmaceuticals still face multiple challenges.

[Ta Kung Pao News]Drug development has always been an expensive and time-consuming task, with an average estimated cost of US$1 billion and 10 to 15 years. With the development of technology, a report released by consulting firm McKinsey at the beginning of this year stated that artificial intelligence is a “once-in-a-century opportunity” for the pharmaceutical industry. AI technology helps speed up drug discovery, approval and listing, and can create economic benefits of US$60 billion to US$110 billion (approximately HK$468 billion to HK$858 billion) for the pharmaceutical industry each year.

Investors are flocking to AI biotech startups. Chip giant Nvidia invested $50 million in biotech company Recursion in 2023 to accelerate the development of artificial intelligence models for drug discovery. Nvidia also developed BioNeMo, a cloud service for generative AI in biology that provides a variety of AI models for small molecules and proteins.

Costs can be reduced to one tenth

New drug development usually goes through multiple stages, including drug design, optimization screening, preclinical research, and clinical trials. Among them, artificial intelligence is expected to play a big role in drug design and optimization screening, especially in the field of rare diseases and orphan drugs, where AI technology is expected to bring new treatment hope to patients.

With the help of artificial intelligence, scientists can design increasingly complex small molecule structures, thereby promoting innovation in small molecule drugs, which has become one of the important directions of AI drug manufacturing. Small molecule drugs are chemically synthesized drugs. Most of the drugs on the market are small molecule drugs, such as aspirin. Based on existing drug data, AI systems use machine learning to more accurately simulate the interactions between biological molecules, quickly screen targets (sites where drugs work), speed up the prediction of potentially effective ingredients, perform pairing and synthetic feasibility analysis, and quickly find more effective potential drug molecular structures. For example, Insilico Medicine used AI to develop a small molecule drug for idiopathic pulmonary fibrosis. Traditionally, this process takes six years and costs more than $400 million, but using generative AI, the company reduced the cost to one-tenth and the time to two and a half years. In addition to small molecule drugs, AI technology is also increasingly being used in the research and development of large molecule drugs such as antibodies and gene therapy.

Clinical safety remains key

However, AI drug development is not easy, and the risks faced by traditional drugs are similar, including issues such as drug effectiveness, safety, and drug resistance. Generative AI used for drug development is trained with precise scientific data, and the possibility of “AI hallucinations” in the system is much lower than that of ordinary chatbots such as ChatGPT. However, before any potential drug is approved for use in patients, it must undergo clinical trials. Even if AI can speed up the process, it cannot avoid the clinical “trial and error” process, which may take several more years.

Many candidate drugs that work well in the laboratory ultimately fail when tested in humans. According to statistics, nearly 90% of candidate drugs that enter clinical human trials end in failure, usually due to lack of efficacy or unforeseen side effects. In the past year, the first batch of drugs designed with the participation of artificial intelligence did not progress as expected. Some were directly suspended from research and development, and some were reduced in clinical trial priority, proving that clinical risks and challenges are huge.

Richard Law, a biologist and chief business officer of Exscientia, an AI drug development company, said that the entire process of drug discovery is inseparable from failure. “The cost of developing a drug is very high because you have to design and test 20 drugs before you can make one drug work.”

AI drug production relies on data and algorithms, and the problems of data bias and algorithm security still need to be solved. Training AI is expensive and consumes a large number of high-performance chips. Talents who master AI technology and have a pharmaceutical background are extremely scarce. Another problem facing AI drug production is regulation. The U.S. Food and Drug Administration (FDA) revealed last year that the number of drug applications that combine artificial intelligence and machine learning elements has increased dramatically in the past five years. The FDA has not yet issued specific regulatory approval guidelines. (Comprehensive report)