The research and development of new drugs has always been a complex project that consumes time and money, but now, scientists seem to have found a way to solve it-the introduction of artificial intelligence technology.
According to a report by MIT Technology Review magazine on September 3, in a new drug development that introduced AI technology, a team from the artificial intelligence pharmaceutical startup Insilicon Medicine collaborated with scientists from the University of Toronto to develop It only took 46 days from starting to develop a new targeted drug to completing preliminary biological verification. The findings were published this week in the journal Nature Biotechnology.
This landmark study confirms that AI technology can help accelerate drug development, which means that patent protection period is extended, thus improving the economics of drug development. If this method can be generalized, it will be widely adopted by the pharmaceutical industry.
AI assisted in shortening the 8-year drug development time to 46 days
Based on two popular artificial intelligence technologies, generative adversarial networks and reinforcement learning, the team developed this drug A new AI system - Generative Tension Reinforcement Learning (GENTRL) - was introduced.
The researchers chose to target DDR1 kinase, a tyrosine kinase expressed in epithelial cells, a protein closely related to tissue fibrotic diseases. After identifying the target, the GENTRL system spent 21 days designing 30,000 different molecular structures. Then, by reviewing known molecules that work on drug targets in previous research and patents, it prioritized new ones that could be synthesized in the laboratory. molecular structure.
The research and development results titled "Deep Learning Can Rapidly Identify Effective DDR1 Kinase Inhibitors" have been published in the journal "Nature Biotechnology". Screenshot from "Nature Biotechnology" magazine
Among the 6 candidate DDR1 inhibitor compounds designed and synthesized by GENTRL, 4 compounds were active in biochemical analysis. In the next stage of in vitro cell experiments, 2 of the 4 active compounds demonstrated the expected DDR1 inhibitory ability and could effectively reduce the levels of markers related to the fibrosis process. Through comparison, the most promising compound was further successfully verified in in vivo experiments in mice.
From the initial target identification, molecular structure screening of potential new drugs, synthesis to preclinical biological verification, the GENTRL system shortens the work that takes at least 8 years to complete with traditional drug development methods to only 46 days .
Michael Levitt, winner of the 2013 Nobel Prize in Chemistry and professor of structural biology at Stanford University, commented, "This paper is certainly an impressive advance and is likely to be applicable to many other aspects of drug design. Question. Based on state-of-the-art reinforcement learning, I was also impressed by the breadth of the research as it relates to molecular modeling, affinity measurements, and animal studies. ”
AI replaces medicinal chemist role. Becoming mainstream
"MIT Technology Review" magazine pointed out that this landmark research may change the "cost-consuming, time-consuming and labor-intensive" dilemma faced by new drug research and development.
This landmark study may change the dilemma faced by new drug research and development. According to the "MIT Technology Review" magazine
"Artificial intelligence will have a revolutionary impact on the pharmaceutical industry, and we need more experimental verification results to accelerate this progress," said many core players in the field of artificial intelligence. Jürgen Schmidhuber, inventor of the technology and initial concept and professor at the Swiss Institute of Artificial Intelligence, said.
As we all know, bringing a new drug to the market costs a lot of money and time. According to data from the Tufts Center for the Study of drug Development, a new drug starts from It can take 10 years from the beginning of research and development to the final launch, costing up to $2.6 billion, and the vast majority of drug candidates will fail during the testing stage.
Therefore, reducing the R&D cycle and economic costs is crucial to the success of drug R&D activities in the pharmaceutical field.
According to "Forbes" magazine, using Insilicon Medicine's method, the development cost of this drug was only US$150,000.
Insilicon Medicine hopes to bring AI deep learning into the drug development process. According to "Forbes" magazine
Charles Cantor, chief scientist of the U.S. Department of Energy's Human Genome Project and a professor at Boston University, said that there are many exaggerations about the prospects of artificial intelligence in improving health care and developing new medical tools. statement. However, the results, recently published in the journal Nature Biotechnology, do have important implications.
It demonstrates firstly that artificial intelligence can replace a role usually played by medicinal chemists, which is often understaffed; secondly, the accelerated speed of drug development means that patent protection periods are extended, thereby improving Economics of drug development. "If this method can be generalized, it will become a widely adopted method in the pharmaceutical industry," said Dr. Cantor.
Of course, this is only the first step in global drug research and development efforts. Although this is a milestone showing the potential of artificial intelligence to identify drug candidates, it will still take years of clinical trials and millions of dollars of research before any promising drugs are approved for treatment.
AI technology can quickly identify effective DDR1 kinase inhibitors. According to Insilicon Medicine
“This paper is an important milestone in our road to artificial intelligence-driven drug development. We started working on AI synthetic chemistry in 2015, but when Insilicon’s theoretical paper was published in 2016 Everyone was skeptical at the time of publication, and now that this technology is becoming mainstream, we are excited that it is being validated in animal experiments and that these models are suitable when integrated into a full-scale drug development process. Numerous targets. We are collaborating with leading biotech companies to further push the limits of synthetic chemistry and synthetic biology," said Dr. Alex Zhavoronkov, first author of the paper and founder and CEO of Insilico Medicine.
Editor Lu Yanfei