2023-08-10 10:45:26
The key to gaining insight into proteins, such as those that govern cancer, COVID-19, and other diseases, is quite simple. It is enough to identify their chemical structure and find what other proteins can bind to them. The difficult part is putting it into practice. The potential combinations of component positions in molecules are more numerous than the estimated number of atoms that exist in the universe. Finding out exactly how a certain molecular process occurs is therefore an enormously difficult task.
To speed up this task a bit, a team including Nathaniel R. Bennett and Brian Coventry, both from the University of Washington in Seattle, have turned to deep learning techniques (a form of artificial intelligence used by systems like ChatGPT or DALL-E). The result has been to multiply by 10 the rate of correct checks carried out in the laboratory.
Deep learning uses computer algorithms to find and analyze patterns in data and make inferences from those patterns.
Bennett and his colleagues used deep learning techniques to obtain revealing structural details about molecular processes of interest, producing theoretical models that, when tested, turned out to be highly accurate.
Bennett’s team enlisted the help of the Frontera supercomputer, located at the Texas Advanced Computing Center.
The NSF-contributed Frontier supercomputer at the Texas Advanced Computing Center. (Photo: TACC)
With the innovations introduced, work was done more than 200 times faster than the best previous software.
Bennett and his colleagues discuss the technical details of their new approach in the academic journal Nature Communications, under the title “Improving de novo protein binder design with deep learning.” (Source: NCYT from Amazings)
#Artificial #intelligence #investigate #proteins