In living organisms, proteins perform important roles, and their function is closely associated with their structure. Reliable prediction of protein function by computational means has become crucial due to the growing difference between the number of proteins being discovered and their functional characterization (in particular as a result of experimental limitations). Scientists can create high-resolution, three-dimensional images of tiny molecules such as proteins via cryo-electron microscopy (cryo-EM). This approach works best for imaging proteins that exist in only one conformation, but a machine-learning algorithm has now been developed by MIT researchers that let them recognise many potential structures that a protein can take. Protein structure can also be experimentally determined using cryo-EM, which produces hundreds of thousands, or even millions, of two-dimensional images of protein samples frozen in a thin layer of ice, unlike AI techniques that attempt to predict protein structure from sequence data alone. In a process called reconstruction, computer algorithms then piece together these images, taken from various angles, into a three-dimensional representation of the protein. A new AI-based program for the reconstruction of multiple structures and motions of the imaged protein, a major objective in the protein science community, is stated by the researchers. Instead of using the conventional representation of protein structure on a 3D lattice as electron-scattering intensities, which is inefficient for modeling multiple structures, the researchers implemented a new architecture of the neural network that can effectively produce in a single model the full collection of structures.

Visualizing a method with several phases: By studying structures that form during the process of assembling ribosomes—the cell organelles responsible for reading messenger RNA and translating it into proteins, the researchers showed the usefulness of their new approach. While a postdoc at the Scripps Research Institute, Davis started researching the structure of ribosomes. There are two main subunits of ribosomes, each containing several individual proteins which are assembled in a multistep process. Davis stalled the process at various points in order to research the steps of ribosome assembly in depth, and then took electron microscope images of the resulting structures. At certain stages, the blocking of the assembly resulted in only a single structure being accumulated, indicating that there is only one way to take that step. Blocking other stages, however, resulted in several different structures, indicating that in a number of ways the assembly could occur.

Media Contact:

Allison Grey

Journal Manager

Journal of Clinical chemistry and Laboratory Medicine

Email: clinchem@longdomjournal.org