Artificial intelligence is liberating the hands of chemists. In the future, we just need to "prescribe medicine"
Recently, Xiangshan Science Conference focused on green ecology and chemical engineering. In addition to various cutting-edge chemical engineering technologies, the application of artificial intelligence in this field has become a new topic of concern.
"Too many variables in the process of synthetic chemistry really make it more difficult to prepare flux and screen flux reaction conditions. But in the era of big data and artificial intelligence, this situation will change greatly. " Ma Dawei, deputy director of Shanghai Institute of Organic Chemistry, Chinese Academy of Sciences, introduced at the meeting that artificial intelligence is expected to help synthetic chemistry research to a new level.
"Labor-intensive" will become a thing of the past.
"In the field of organic synthesis, synthetic chemists claimed to give enough students and financial support in the last century to synthesize any complex organic small molecules. But until today, whether it is the development of new drugs or new materials, people have found that the decisive step of the process is often the synthesis of compounds. " Ma Dawei introduced.
Ma Dawei explained that this is because, although most of the evaluation of the properties of compounds has entered the flux screening stage, chemical synthesis still stays in the manual operation stage.
In the field of chemical synthesis, due to the infinite changes under different chemical reaction conditions, researchers basically rely on manual design of chemical synthesis routes — — Especially in the field of drug molecular synthesis. This makes chemical synthesis more like a "labor-intensive" job.
But in the era of artificial intelligence, this situation will gradually become a thing of the past. Computer programs based on big data and artificial intelligence are becoming more and more mature and practical in assisting researchers to design chemical synthesis routes.
In April last year, a German research team published a paper in the journal Nature, saying that they could conduct chemical synthesis analysis at an unprecedented rate with artificial intelligence systems, which would greatly improve the efficiency of researchers in developing new drugs and other compounds.
According to the content of this paper, it takes only 5 seconds for researchers to design a synthetic route of a molecule by using artificial intelligence system. Ma Dawei lamented that if you design it manually, you won’t be able to see the molecular structure clearly in 5 seconds, and you may not be able to design the molecule in 5 minutes.
Li Chenghui, an associate professor at the School of Chemistry and Chemical Engineering of Nanjing University, gained a lot of inspiration after learning about the synthetic route of molecules recommended by artificial intelligence algorithms. He is also considering how to use artificial intelligence to help him do chemical research.
Li Chenghui recently discovered a new intramolecular cyclization reaction, hoping to know whether this reaction also exists in other molecules. According to the previous research methods, his workload will be very heavy. Because to test whether this reaction is universal, a lot of experiments need to be done with different molecules. "If you have the help of artificial intelligence, you can do it in a targeted manner." Li Chenghui said.
Start playing the right-hand man
"Small flux reaction condition screening equipment has become a conventional weapon in some companies and universities, and large-scale reaction matrix robots with large flux have been born." Ma Dawei believes that the popularization of these devices may bring great changes to the future research of synthetic chemistry, which will largely liberate the hands of synthetic chemists and allow researchers to spend more time on data analysis and top-level design of reactions.
In an interview with the reporter of Science and Technology Daily, Ma Dawei said that the artificial intelligence program used to predict the synthesis route of ordinary chemical small molecules has become mature. Even some complex organic small molecules, such as drug molecules, have not many synthetic reactions because of their single structural units, and artificial intelligence programs can directly assist in designing synthetic routes. This will be of great help to junior researchers.
"Now doing drug synthesis research, if you know the structure of a drug molecule, the artificial intelligence system can analyze many synthetic routes according to its structure and recommend the best route." Li Chenghui said that in the past, such research needed very experienced organic chemistry experts, but with the help of artificial intelligence, researchers engaged in inorganic chemistry can also do it in the future.
However, when synthesizing target molecules with complex structures, such as complex molecules existing in the natural world that need more than 20 steps to synthesize, artificial intelligence programs can predict many different routes.
"Researchers may only be able to think of more than 20 synthetic routes, and these programs may provide hundreds of possible synthetic routes, which is very helpful for researchers, because some routes may never be thought of." Ma Dawei said.
Ma Dawei told the Science and Technology Daily reporter that according to the predicted routes provided by artificial intelligence programs, experienced researchers can judge which routes are good and which are suggestive, and then help researchers think about some problems. This can provide more opportunities for chemical synthesis. Because according to the tips of artificial intelligence program and further deep thinking of human brain, a better chemical synthesis route may be designed.
In the future, just "prescribe according to the medicine"
Ma Dawei believes that the above changes are of great significance to chemical synthesis. Because the artificial intelligence program may be like a "fool machine" in the future, researchers only need to "prescribe medicine according to its recommendation" when conducting chemical synthesis.
In the past, understanding whether a chemical reaction condition is feasible was a process of trial and error. It needs manual testing one by one, which involves a lot of repetitive work. Nowadays, artificial intelligence can be used to quickly detect chemical reaction data, and then the reaction conditions of flux can be tested on the machine. The whole system can test thousands of reaction conditions a day. For an artificial person, it is good to do 20 chemical reaction tests every day in a laboratory.
Ma Dawei told the Science and Technology Daily reporter that some large pharmaceutical companies have used artificial intelligence to help design chemical synthesis routes and become the first users of such technologies. In addition to drug research and development, new material research and development and other fields including compound synthesis can use artificial intelligence to recommend chemical synthesis routes.
"In the future, synthesizing molecules that didn’t exist in the past is like building a house. Let artificial intelligence draw the blueprint first, and researchers just need to build it according to the blueprint." Ma Dawei said.
Li Chenghui also believes that artificial intelligence will be promising in the field of chemistry. In his view, the application of artificial intelligence in the future chemical field will be like an artificial intelligence doctor, which can master and digest massive synthesis methods, synthesis routes, material structures and properties, etc. When researchers do chemical synthesis research, it can help analyze and solve many problems.
But in general, the role played by artificial intelligence is still auxiliary. It can liberate chemical researchers from heavy manual labor and provide some reference for their research. But the recommended results also need researchers to use professional knowledge and experience to judge which ones are really feasible.
In addition, the chemical synthesis route recommended by artificial intelligence is still in the model machine stage. Ma Dawei believes that chemical researchers need to constantly provide some innovative strategies for artificial intelligence and promote the continuous optimization of artificial intelligence systems, which in turn will promote chemical researchers to conduct more in-depth research.
Related links
Artificial intelligence robot helps to find new molecules quickly.
As early as last year, chemists at the University of Glasgow in the United Kingdom had discussed how to train an artificial intelligence organic chemical synthesis robot to automatically explore a large number of chemical reactions.
This "self-driven" system is based on machine learning algorithm, which can discover new reactions and molecules, and allows new molecules of interest to be located by digital chemical data-driven methods, instead of being limited to a known database and conventional organic synthesis rules. The result may reduce the cost of discovering new drug molecules, new chemical products (including materials), polymers and molecules used in high-tech applications (such as imaging).
The research team demonstrated the potential of the system by using a combination of 18 different starting chemicals to simulate about 1000 reactions. After exploring about 100 possible reactions, the robot can predict with more than 80% accuracy which initial chemical combinations should be explored to produce new reactions and molecules. By exploring these reactions, they discovered a series of new molecules and reactions that were previously unknown. The researchers found four new reactions, one of which was classified as the top 1% of the most unique known reactions.
The researchers said that this method is a key step in the digitalization of chemistry, which will allow real-time retrieval of chemical space, thus helping to discover new drugs, cut costs, save time, improve safety, reduce waste, and help chemistry enter a new digital era. (Reporter Liu Yuanyuan)