- Статьи
- Science and technology
- An infectious example: scientists have come close to creating 56 new antibiotics
An infectious example: scientists have come close to creating 56 new antibiotics
Russian scientists have developed an artificial intelligence-based system that automatically finds candidate substances for the role of antibiotics. AI has already proposed 56 compounds that may be effective against drug-resistant strains of E. coli. Now the specialists will have to test the properties of the identified molecules in practice. Although not all of them will reach real use, experts believe that the development will reduce the time needed to create new drugs by months, and possibly even years. The creators of the system believe that in the future the algorithm can be adapted to search for antiviral and antifungal agents.
AI to search for drug molecules
ITMO scientists have created an algorithm based on artificial intelligence that automatically selects candidate molecules for the creation of new antibiotics. The system is looking for compounds to which the bacteria will not be able to quickly develop resistance.
The development has already been tested in action and 56 new compounds have been discovered with its help, which can form the basis of more effective drugs against E. coli. This bacterium often causes severe gastrointestinal infections.
— Research in the laboratory will help experimentally verify the actual activity of the compounds obtained. Usually, of the many candidate molecules discovered by computational methods, one or two are suitable in practice. If the results show the high efficiency of our compounds, then we can consider a patent. We are currently looking for just such a laboratory," Anastasia Orlova, one of the authors of the study, an engineer at ITMO's Advanced Engineering School, and a graduate student, told Izvestia.
Antibiotics block or modify the action of proteins of harmful microorganisms and inhibit their growth and reproduction. But over time, the drugs lose their effectiveness, as the bacteria evolve and change their proteins so that the drugs do not act on them. This process is called antibiotic resistance. To overcome it, experts all over the world are looking for substances to which microbes have not yet had time to adapt. However, sooner or later, pathogens will learn to bypass the effects of new drugs, so the race between pharmacists and infections continues continuously.
To find new antibiotics, scientists use high-throughput screening. Within its framework, specialists check molecules from databases for compliance with criteria such as the absence of toxicity, the ability to bind to target proteins and the possibility of laboratory synthesis, the developers explained. As a rule, such a process can take up to several weeks and does not allow the creation of fundamentally new compounds.
Machine learning helps in solving this problem, but its methods are still imperfect. Many existing algorithms generate active molecules for only one protein, which is often not enough to create an effective antibiotic to which bacteria cannot quickly develop resistance.
From a molecule to a medicine
The new algorithm searches for substances that have activity against two proteins at the same time. To combat resistant strains of E. coli, he proposed 56 benzimidazole-based compounds. There are few drugs with such a component on the market yet. Unlike analogues, the new system can generate molecules at once, taking into account a variety of properties, ensuring their synthesizability. AI also takes into account the absence of toxicity and side effects, binding to target proteins, similarity to other drugs, and biological activity, the developers noted.
According to them, experiments have shown that the accuracy of calculations reaches 81%. Experts have already studied the generated compounds using computer modeling, which revealed that some of them have higher activity than the novobiocin antibiotic registered on the market.
The developed technology will make it easier and faster to create medicines for months, and possibly even years, the authors of the project believe. In the future, the algorithm can be adapted to search for drugs against other bacteria, for example, salmonella, klebsiella, meningococcus, as well as to create antiviral and antifungal agents.
Modeling the interaction of molecules, that is, the in silico experiment, is the most important achievement in modern pharmacology and biology in general, explained Mikhail Bolkov, a researcher at the Institute for the Study of Aging at the Russian Gerontological Research and Clinical Center at Pirogov University.
— For example, why try to mix thousands of compounds and observe them under the same experimental conditions, trying to find those that react in the right way, if the laws of chemistry and physics are known by which this happens? By simulating their interaction in the program, scientists avoid wasting resources, saving months and even years of time on experiments. And only for sure the working compounds will already go into a real experiment to test their interaction in a cell, then in animals and finally in humans," he told Izvestia.
The project demonstrates the right direction for the digital transformation of the pharmaceutical industry — from empirical search to a data analysis approach. With successful laboratory confirmation, the technology can become the basis for creating a biotech startup with high economic potential, says Maxim Kolyasnikov, associate professor at the UrFU Institute of Economics and Management.
The use of modern computing tools, including AI, really makes it possible to accelerate and reduce the cost of drug development. However, according to Stanislav Stragnov, head of the master's program "Applied Analysis in the Medical Field" at MIPT, the main problem is not the creation of a candidate compound, but obtaining a drug that has proven its safety and effectiveness during preclinical and clinical trials.
The expert noted that 90% of theoretically promising drugs do not go this way. It is premature to celebrate success until a laboratory test is carried out, he concluded.
Переведено сервисом «Яндекс Переводчик»
