Breakthrough in Protein Origami
An important topic in my biochemistry studies were proteins and their spatial structure. These molecular tools perform indispensable services in all living beings, and their functions depend on the concrete individual shape of the macromolecule. For a long time, therefore, one of the major problems in the life sciences was: How can one deduce from the sequence of amino acids of a protein encoded in the genome what three-dimensional shape the molecule folds into? The former was easily elucidated in the 1990s, but determining the structure proved to be a tough experimental chunk. Either the proteins had to be crystallized at great expense for an X-ray structure analysis, or an attempt was made to determine how the individual sections of a protein are spatially related to one another using the then new approach of nuclear magnetic resonance. I still remember Kurt Wüthrich's lecture on the latter method at ETH Zurich, in which the later Nobel Prize winner explained the difficulties involved.
Now the dream of biochemists and molecular biologists, which has been cherished for decades, has apparently come true - thanks to artificial intelligence. The "AlphaFold 2" algorithm from DeepMind can now calculate the spatial shape of proteins from their amino acid sequence with astonishing precision, as biophysicist Gunnar Schröder from Forschungszentrum Jülich explains from p. 46 onwards. And all this without complex laboratory tests and in a manageable time.
At the end of July, DeepMind, together with the European Molecular Biology Laboratory EMBL in Heidelberg, even reported that they had now created the structures for almost all human proteins and those of various important model organisms in this way and made them openly accessible for research. The reliability of the determinations still varies, but is already in a very high range for the majority of the proteins. By the end of the year, the number of published structures is expected to increase from around 365,000 to around 130 million: That would be almost half of all known proteins!
This could be another golden age for the life sciences, from which we will all benefit. Because knowledge of the structures of the relevant target proteins is often a crucial prerequisite for the development of effective drugs against threatening diseases. Accordingly, the article also opens our new three-part series "From molecular structure to drug".
Have fun reading, Hartwig Hanser