Google’s AI helps determine the most complex protein knots so far

Most complex protein knots.


A particularly fascinating phenomenon arises for proteins that contain a topological knot in their polypeptide backbone, that is, proteins that would not fully disentangle after being pulled from both ends. Only about 20 different protein families containing knots have been identified in the past two decades.

Nevertheless, knotted proteins pose a challenge to protein folding and evolution. Since it contributes substantially to, among other things, the understanding of many diseases and their treatment, this information regarding the so-called “folding” of proteins is in high demand. These factors led to the creation of AlphaFold, artificial intelligence that predicts 3D structures.

Using this AI, a research team including researchers from Johannes Gutenberg University Mainz (JGU) and the University of California, Los Angeles, has closely observed these structures and examined them with respect to knots.

Maarten A. Brems, a Ph.D. student in the group of Dr. Peter Virnau at Mainz University, said, “We investigated numerically all – that is some 100,000 – predictions of AlphaFold for new protein knots. The goal was to identify rare, high-quality structures containing complex and previously unknown protein knots to provide a basis for experimental verification of AlphaFold’s predictions. The study discovered the most complex knotted protein to date and the first composite knots in proteins. The latter can be thought of as two separate knots on the same string.”

Robert Runkel, a theoretical physicist, involved in the project, said, “These discoveries also provide insight into the evolutionary mechanisms behind such rare proteins.”

Dr. Peter Virnau is pleased with the results“We have already established a collaboration with our colleague Dr. Todd Yeates from UCLA to confirm these structures experimentally. This line of research will shape the biophysics community’s view of artificial intelligence – and we are fortunate to have an expert like Todd Yeates involved.”

Journal References:

  1. M. Brems et al., AlphaFold predicts the most complex protein knot and composite protein knots Protein, Protein Science 31: 8, 13 July 2022. DOI: 10.1002/pro.4380
  2. J. Jumper et al., Highly accurate protein structure prediction with AlphaFold, Nature 596, 583-589, 15 July 2021. DOI: 10.1038/s41586-021-03819-2
Latest Updates