Using a algorithm to synthesize new drugs, a peak matchup is about to open.
[ China Pharmaceutical Network Technology News ] Chemists predict that humans can synthesize 1060 kinds of drug molecules, which are smaller than the atoms in the solar system. They hope that the algorithms in the chip can classify hundreds of silicon oxides to help researchers quickly and reasonably find the best candidate. Many supporters believe that some options make drugs safer, reduce the probability of drugs failing in clinical trials, and discover new treatments, perhaps they can help open unknown areas.
One molecule is a world.
This is true for drug developers. In this "Nature" article, Ottawa reporter Asher Mullard will pull you into a world of medicinal herbs.
The story is about to start -
In 2016, Sunovion, the US subsidiary of Sumitomo Pharmaceuticals, convened several experienced developers to assign them an unusual task:
Let's have a game and see who can find the best new medicine.
In front of the researchers are hundreds of chemical structures, of which less than one-tenth of the structure is labeled with biological effects. Researchers need to select the most appropriate of the many structures and synthesize drugs based on their chemical and biological knowledge.
Of the 11 contestants, 10 struggled for hours, but one contestant easily reached the results in just a few milliseconds.
It is an algorithm.
As the head of the chemical informatics department of the startup Exscientia, Willem van Hoorn designed the computer program. At the time, Exscientia was a startup based in Dundee, England, which studied AI design drugs and looked forward to working with Sunovion.
But Sunovion didn't take this small startup into the eye. “In the beginning, my credibility was in jeopardy,†van Hoorn said. After a score of more than 20 game rounds, he finally breathed a sigh of relief. This algorithm seems to have mastered some chemical "black magic" and defeated everyone except a synthetic expert.
This "game" opened a milestone in the collaboration between Sunovion and Exscientia. "It was this game that convinced the researchers," recalls Scott Brown, director of computational chemistry at Sunovion.
Except for academia and industry, and the desire to use computer technology to explore the world of the mysterious chemical world continues to grow, and Exscientia is only one of them.
Chemists predict that humans can synthesize 1060 kinds of drug molecules, which are smaller than the atoms in the solar system. They hope that the algorithms in the chip can classify hundreds of silicon oxides to help researchers quickly and reasonably find the best candidate.
Many supporters believe that these options make drugs safer, reduce the probability of drugs failing in clinical trials, and discover new treatments, perhaps they can help open unknown areas.
Of course, there are also many voices of opposition. Many drug chemists are still skeptical about this "hype" behavior, and do not believe that complex and wonderful chemical reactions can eventually be summarized in a few lines of code. Some AI supporters also admit that many attempts fail to achieve the desired results. Computer-generated compounds are sometimes difficult to synthesize. Sometimes the reaction conditions are too harsh and dangerous to synthesize.
“Computer-generated synthesis methods are not very viable when researchers don’t understand the field,†says van Hoorn. At the same time, he said that human experts can train these unassuming digital designers. “I think some ideas can be implemented with the intervention of computer scientists and chemists,†van Hoorn said.
space exploration
Want to travel in the chemical universe, there is a "map" is essential. In 2001, Jean-Louis Reymond, a chemist at the University of Bern, Switzerland, began to use his computer to draw chemical space as much as possible. Sixteen years later, he has compiled the world's largest small molecule database GDB-17. This is a huge compound virtual database, with a total of 17 atoms involved in the study, forming a total of 166 billion combinations, just the number of Reymond's computers. "Now, it takes only 10 hours to compile a list of compounds in a database with a single computer," Reymond said.
In order to clarify the starting point of the synthesis of these large amounts of drugs, Reymond thought of a way to piece together his chemical universe. He draws inspiration from the periodic table of elements, gathering compounds in a multidimensional space, and compounds with similar structures are also tied together. The position of the compound is assigned based on 42 characteristics, such as the number of carbon atoms in each molecule.
Each of the drugs entering the market has millions of structurally similar compounds that may be just a change in the position of a hydrogen bond or a double bond. Some of them may work better than drugs entering the market, but chemists cannot fully conceive of all isomers.
"You can't use a pen and a piece of paper to deal with these isomers," Reymond said.
There is a way to help solve this problem, and Reymond and his team can identify "proximities" with therapeutic potential by looking for similarities between compounds. Using a special drug as the starting point, the researchers combed all of the 166 billion compounds in the database in 3 minutes to screen out the best alternatives for synthetic drugs.
In a proof-of-concept experiment, Reymond began to structure a molecule with a known structure that binds to the nicotine acetylcholine receptor. This is a way to distort the nervous system and muscles, and for this purpose they have compiled an alternate list of 344 related compounds.
Later, the team synthesized three of these structures and found that two of them can effectively activate the receptor and also treat muscle atrophy in the elderly. Reymond said that this method is like finding a gold mine through a geological map. “You need to choose the place you want to dig in some way,†he said.
For computers, this means using chips to screen out small molecules that bind to a given protein from the vast structure in the chemical library. First, researchers need to "photograph" proteins with X-rays to determine the binding sites. Then, using the molecular docking algorithm, the computational chemist can find the best match for the anchor point through the set of compounds.
With the explosive growth of computing power, the capabilities of these algorithms have increased. The chemists at the University of California, San Francisco, under the leadership of Brian Shoichet, demonstrated the potential of this approach in 2016 and found a new painkiller.
The Shoichet team found the best by screening more than 3 million commercially available compounds, selectively activating μ-opioid receptor signaling to alleviate pain without disturbing the closely related signaling pathways of beta-arrestin. The researchers quickly found 23 of the most probable compounds from a huge library of compounds.
In addition to the university professor, Shoichet has a dual identity, and he is the co-founder of Epiodyne, a biotechnology company based in San Francisco, California. Epiodyne wants to find safer painkillers in the same way, and they plan to find alternative painkillers that are easy to synthesize from compounds that have never been synthesized before.
Currently, commercial drug development companies are testing this approach, such as the Massachusetts-based biotechnology company Nimbus Therapeutics. It combines virtual compounds on the screen with compounds produced in nature. It's unclear whether this approach will promote the emergence of new drugs, but the company's CEO Don Nicholson said that this is "a good focus at least on drug design projects."
Peak matchup
Although these data search methods have been tried and tested, the computer can only be executed in accordance with script instructions. Machine learning has been at the forefront of discovering new drugs, and algorithms use data and experience to learn which compounds and which targets are combined to find new models. About 12 companies have sprung up and collaborated with large pharmaceutical companies to create drug tracking algorithms.
Andrew Hopkins, CEO of Exscientia, a UK drug development company, has proven that these new methods are reliable. A new drug usually takes 4.5 years from discovery, optimization, to clinical research. Chemists often synthesize thousands of compounds to ensure optimal selection. Even so, the possibility of commercialization of drugs is still minimal. .
Exscientia tries to combine some algorithms (the one mentioned above is amazing Sunovion's algorithm is one of them), the combination of algorithms may be able to reduce the drug development time from 4.5 years to 1 year, and can effectively reduce the number of compounds that need to be considered in the early stage. .
Brandon Allgood, CTO of Numerate, an AI drug design company based in San Bruno, Calif., said the algorithm can also help drug developers decide which compounds to pass away as early as possible.
He believes that if a compound fails in toxicity or absorption tests after a few months, then previous research and testing will become meaningless. Allgood was a cosmologist before starting chemistry with AI. "It takes only a millisecond to control it with AI," Allgood said.
This year, Numerate has entered into two agreements with pharmaceutical companies, including a joint project with Suresnes, France, which wants to treat heart failure and arrhythmias through clinical trials.
Although new funds continue to flow into the entire AI pharmaceutical industry, the correctness of these calculations remains to be proven. Compared to other subject libraries, Reymond's collection is huge, but it covers only a small part of the chemical universe.
Many of the calculated alternative compounds are difficult to synthesize in the laboratory, and chemists have to work hard to find the raw materials for the recommended compounds, which can take months or longer. Even so, there is no guarantee that this molecule will work once it is made.
Reymond's method predicts the activity profile of a compound, but the prediction accuracy is only 5-10%, which means that chemists must choose between more than 20 compounds to find the one that best meets expectations. "I want to say that the bottleneck in exploring the chemical space is to dare to make this compound," Reymond said. To this end, he recently turned the chemical library into a short chain of 10 million molecules that are easy to manufacture but still cover a wide range of properties.
Cambridge University scientist Mark Murcko believes that the focus of computational chemists should not be on algorithmic strategies, but more on improving them to improve data sets. “The best way to predict the model I know is to feed it continuously,†he said.
For the CEO of Exscientia, these collaborations are crucial. This requires computational scientists to spend decades writing programs to defeat chess masters. Then, in 1997, IBM's Deep Blue defeated chess master Kasparov. But these failures do not mean the end of the chess game. Instead, Kasparov created a two-player chess game, with one team and one AI per team.
"The combination of human and AI can outperform any human being and can outperform any algorithm," Hopkins said. He wants the same data processing, creativity and common sense to change the discovery of new drugs. "We are at the peak of the match between Kasparov and Deep Blue," Hopkins said.
Original title: Synthesizing new drugs with an algorithm: a new match between Kasparov and the deep blue peak
Teach you to use APP (Pharmaceutical Pass) to seize business opportunities
One molecule is a world.
This is true for drug developers. In this "Nature" article, Ottawa reporter Asher Mullard will pull you into a world of medicinal herbs.
The story is about to start -
In 2016, Sunovion, the US subsidiary of Sumitomo Pharmaceuticals, convened several experienced developers to assign them an unusual task:
Let's have a game and see who can find the best new medicine.
In front of the researchers are hundreds of chemical structures, of which less than one-tenth of the structure is labeled with biological effects. Researchers need to select the most appropriate of the many structures and synthesize drugs based on their chemical and biological knowledge.
Of the 11 contestants, 10 struggled for hours, but one contestant easily reached the results in just a few milliseconds.
It is an algorithm.
As the head of the chemical informatics department of the startup Exscientia, Willem van Hoorn designed the computer program. At the time, Exscientia was a startup based in Dundee, England, which studied AI design drugs and looked forward to working with Sunovion.
But Sunovion didn't take this small startup into the eye. “In the beginning, my credibility was in jeopardy,†van Hoorn said. After a score of more than 20 game rounds, he finally breathed a sigh of relief. This algorithm seems to have mastered some chemical "black magic" and defeated everyone except a synthetic expert.
This "game" opened a milestone in the collaboration between Sunovion and Exscientia. "It was this game that convinced the researchers," recalls Scott Brown, director of computational chemistry at Sunovion.
Except for academia and industry, and the desire to use computer technology to explore the world of the mysterious chemical world continues to grow, and Exscientia is only one of them.
Chemists predict that humans can synthesize 1060 kinds of drug molecules, which are smaller than the atoms in the solar system. They hope that the algorithms in the chip can classify hundreds of silicon oxides to help researchers quickly and reasonably find the best candidate.
Many supporters believe that these options make drugs safer, reduce the probability of drugs failing in clinical trials, and discover new treatments, perhaps they can help open unknown areas.
Of course, there are also many voices of opposition. Many drug chemists are still skeptical about this "hype" behavior, and do not believe that complex and wonderful chemical reactions can eventually be summarized in a few lines of code. Some AI supporters also admit that many attempts fail to achieve the desired results. Computer-generated compounds are sometimes difficult to synthesize. Sometimes the reaction conditions are too harsh and dangerous to synthesize.
“Computer-generated synthesis methods are not very viable when researchers don’t understand the field,†says van Hoorn. At the same time, he said that human experts can train these unassuming digital designers. “I think some ideas can be implemented with the intervention of computer scientists and chemists,†van Hoorn said.
space exploration
Want to travel in the chemical universe, there is a "map" is essential. In 2001, Jean-Louis Reymond, a chemist at the University of Bern, Switzerland, began to use his computer to draw chemical space as much as possible. Sixteen years later, he has compiled the world's largest small molecule database GDB-17. This is a huge compound virtual database, with a total of 17 atoms involved in the study, forming a total of 166 billion combinations, just the number of Reymond's computers. "Now, it takes only 10 hours to compile a list of compounds in a database with a single computer," Reymond said.
In order to clarify the starting point of the synthesis of these large amounts of drugs, Reymond thought of a way to piece together his chemical universe. He draws inspiration from the periodic table of elements, gathering compounds in a multidimensional space, and compounds with similar structures are also tied together. The position of the compound is assigned based on 42 characteristics, such as the number of carbon atoms in each molecule.
Each of the drugs entering the market has millions of structurally similar compounds that may be just a change in the position of a hydrogen bond or a double bond. Some of them may work better than drugs entering the market, but chemists cannot fully conceive of all isomers.
"You can't use a pen and a piece of paper to deal with these isomers," Reymond said.
There is a way to help solve this problem, and Reymond and his team can identify "proximities" with therapeutic potential by looking for similarities between compounds. Using a special drug as the starting point, the researchers combed all of the 166 billion compounds in the database in 3 minutes to screen out the best alternatives for synthetic drugs.
In a proof-of-concept experiment, Reymond began to structure a molecule with a known structure that binds to the nicotine acetylcholine receptor. This is a way to distort the nervous system and muscles, and for this purpose they have compiled an alternate list of 344 related compounds.
Later, the team synthesized three of these structures and found that two of them can effectively activate the receptor and also treat muscle atrophy in the elderly. Reymond said that this method is like finding a gold mine through a geological map. “You need to choose the place you want to dig in some way,†he said.
For computers, this means using chips to screen out small molecules that bind to a given protein from the vast structure in the chemical library. First, researchers need to "photograph" proteins with X-rays to determine the binding sites. Then, using the molecular docking algorithm, the computational chemist can find the best match for the anchor point through the set of compounds.
With the explosive growth of computing power, the capabilities of these algorithms have increased. The chemists at the University of California, San Francisco, under the leadership of Brian Shoichet, demonstrated the potential of this approach in 2016 and found a new painkiller.
The Shoichet team found the best by screening more than 3 million commercially available compounds, selectively activating μ-opioid receptor signaling to alleviate pain without disturbing the closely related signaling pathways of beta-arrestin. The researchers quickly found 23 of the most probable compounds from a huge library of compounds.
In addition to the university professor, Shoichet has a dual identity, and he is the co-founder of Epiodyne, a biotechnology company based in San Francisco, California. Epiodyne wants to find safer painkillers in the same way, and they plan to find alternative painkillers that are easy to synthesize from compounds that have never been synthesized before.
Currently, commercial drug development companies are testing this approach, such as the Massachusetts-based biotechnology company Nimbus Therapeutics. It combines virtual compounds on the screen with compounds produced in nature. It's unclear whether this approach will promote the emergence of new drugs, but the company's CEO Don Nicholson said that this is "a good focus at least on drug design projects."
Peak matchup
Although these data search methods have been tried and tested, the computer can only be executed in accordance with script instructions. Machine learning has been at the forefront of discovering new drugs, and algorithms use data and experience to learn which compounds and which targets are combined to find new models. About 12 companies have sprung up and collaborated with large pharmaceutical companies to create drug tracking algorithms.
Andrew Hopkins, CEO of Exscientia, a UK drug development company, has proven that these new methods are reliable. A new drug usually takes 4.5 years from discovery, optimization, to clinical research. Chemists often synthesize thousands of compounds to ensure optimal selection. Even so, the possibility of commercialization of drugs is still minimal. .
Exscientia tries to combine some algorithms (the one mentioned above is amazing Sunovion's algorithm is one of them), the combination of algorithms may be able to reduce the drug development time from 4.5 years to 1 year, and can effectively reduce the number of compounds that need to be considered in the early stage. .
Brandon Allgood, CTO of Numerate, an AI drug design company based in San Bruno, Calif., said the algorithm can also help drug developers decide which compounds to pass away as early as possible.
He believes that if a compound fails in toxicity or absorption tests after a few months, then previous research and testing will become meaningless. Allgood was a cosmologist before starting chemistry with AI. "It takes only a millisecond to control it with AI," Allgood said.
This year, Numerate has entered into two agreements with pharmaceutical companies, including a joint project with Suresnes, France, which wants to treat heart failure and arrhythmias through clinical trials.
Although new funds continue to flow into the entire AI pharmaceutical industry, the correctness of these calculations remains to be proven. Compared to other subject libraries, Reymond's collection is huge, but it covers only a small part of the chemical universe.
Many of the calculated alternative compounds are difficult to synthesize in the laboratory, and chemists have to work hard to find the raw materials for the recommended compounds, which can take months or longer. Even so, there is no guarantee that this molecule will work once it is made.
Reymond's method predicts the activity profile of a compound, but the prediction accuracy is only 5-10%, which means that chemists must choose between more than 20 compounds to find the one that best meets expectations. "I want to say that the bottleneck in exploring the chemical space is to dare to make this compound," Reymond said. To this end, he recently turned the chemical library into a short chain of 10 million molecules that are easy to manufacture but still cover a wide range of properties.
Cambridge University scientist Mark Murcko believes that the focus of computational chemists should not be on algorithmic strategies, but more on improving them to improve data sets. “The best way to predict the model I know is to feed it continuously,†he said.
For the CEO of Exscientia, these collaborations are crucial. This requires computational scientists to spend decades writing programs to defeat chess masters. Then, in 1997, IBM's Deep Blue defeated chess master Kasparov. But these failures do not mean the end of the chess game. Instead, Kasparov created a two-player chess game, with one team and one AI per team.
"The combination of human and AI can outperform any human being and can outperform any algorithm," Hopkins said. He wants the same data processing, creativity and common sense to change the discovery of new drugs. "We are at the peak of the match between Kasparov and Deep Blue," Hopkins said.
Original title: Synthesizing new drugs with an algorithm: a new match between Kasparov and the deep blue peak
Teach you to use APP (Pharmaceutical Pass) to seize business opportunities
Pe Medical Tape,Breathable Medical Tape,Medical Surgical Tape,Nonwoven Medical Tape
RFX+CARE Manufacturing Co.,Ltd. , https://www.rk-rfxcare.com