AI enables unprecedented insights into how biomolecules work

A new analytical technique is able to provide hitherto unattainable insights into the extremely rapid dynamics of biomolecules. The team of developers, led by Abbas Ourmazd from the University of Wisconsin–Milwaukee and Robin Santra from DESY, is presenting its clever combination of quantum physics and molecular biology in the scientific journal Nature. The scientists used the technique to track the way in which the photoactive yellow protein (PYP) undergoes changes in its structure in less than a trillionth of a second after being excited by light.

“In order to precisely understand biochemical processes in nature, such as photosynthesis in certain bacteria, it is important to know the detailed sequence of events,” Santra explains their underlying motivation. “When light strikes photoactive proteins, their spatial structure is altered, and this structural change determines what role a protein takes on in nature.” Until now, however, it has been almost impossible to track the exact sequence in which structural changes occur. Only the initial and final states of a molecule before and after a reaction can be determined and interpreted in theoretical terms. “But we don’t know exactly how the energy and shape changes in between the two,” says Santra. 

A peculiar feature of photoactive proteins needs to be taken into consideration: the incident light excites their electron shell to enter a higher quantum state, and this causes an initial change in the shape of the molecule. This change in shape can in turn result in the excited and ground quantum states overlapping each other. In the resulting quantum jump, the excited state reverts to the ground state, whereby the shape of the molecule initially remains unchanged. The conical intersection between the quantum states therefore opens a pathway to a new spatial structure of the protein in the quantum mechanical ground state.

6000 dimensions

“The photoactive yellow protein we studied consists of some 2000 atoms,” explains Santra, who is a Lead Scientist at DESY and a professor of physics at Universität Hamburg. “Since every atom is basically free to move in all three spatial dimensions, there are a total of 6000 options for movement. That leads to a quantum mechanical equation with 6000 dimensions – which even the most powerful computers today are unable to solve.”

However, computer analyses based on machine learning were able to identify patterns in the collective movement of the atoms in the complex molecule. “It’s like when a hand moves: there, too, we don’t look at each atom individually, but at their collective movement,” explains Santra. Unlike a hand, where the possibilities for collective movement are obvious, these options are not as easy to identify in the atoms of a molecule. However, using this technique, the computer was able to reduce the approximately 6000 dimensions to four. By demonstrating this new method, Santra’s team was also able to characterise a conical intersection of quantum states in a complex molecule made up of thousands of atoms for the first time.

Source and original article: Science Business
Read the original publication by DESY on November 3.