2023-11-29 10:45:12
A new active ingredient can be the basis of a new and effective medicine, but to ensure the viability of its manufacturing on an industrial scale, the ideal way to produce it must be found. This challenge is added to the previous one of finding the chemical substance in question. To do this, chemists typically resort to trial and error: they deduce possible laboratory synthesis methods from known chemical reactions and then test each of them with experiments. This research strategy is time-consuming and riddled with dead ends.
Now, a team including, among others, Gisbert Schneider of the Swiss Federal Institute of Technology in Zurich (ETH), and David F. Nippa of the well-known pharmaceutical company Roche, has devised an approach based on artificial intelligence that helps determine the best synthesis method, including its probability of success.
Basically, the researchers trained an artificial intelligence with detailed descriptions of certain chemical reactions taken from scientific literature and others from an automated laboratory of the Roche company. In this laboratory, numerous chemical reactions are carried out on a milligram scale and analyzed simultaneously.
“Combining laboratory automation with artificial intelligence has enormous potential to greatly increase efficiency in chemical synthesis while improving sustainability,” says Nippa.
The predictive capacity of the model generated from this data set was checked, and the result was good in the vast majority of cases.
The new artificial intelligence system is capable of determining the best synthesis method to produce the substance of interest, and also calculates its probability of success. (Images: ETH Zürich)
Interestingly, the predictions improved even further when three-dimensional information about the starting materials was included rather than limiting the information to their two-dimensional chemical formulas. “It seems that the artificial intelligence model develops a kind of three-dimensional chemical understanding,” says Kenneth Atz of the research and development team.
The success rate of the predictions has so impressed the Roche company that it is already using this artificial intelligence to more quickly develop new and more effective variants of known active pharmaceutical ingredients.
Schneider, Nippa and their colleagues present the technical details of their work with the artificial intelligence model in the academic journal Nature Chemistry, under the title “Enabling Late-Stage Drug Diversification by High-Throughput Experimentation with Geometric Deep Learning.” (Source: NCYT from Amazings)
#Artificial #intelligence #find #drugs