Structural biology enters a new era: two prediction algorithms join forces to board "Science"

In December 2020, Google's DeepMind company announced a sensational development: their artificial intelligence (AI) algorithm AlphaFold solved the protein folding problem that has existed for nearly half a century, and accurately predicted the three-dimensional protein based on the amino acid sequence.

In July this year, the details of this research were published in the journal Nature

In a "Science" paper just launched today, these two state-of-the-art protein prediction tools have joined forces to predict the three-dimensional structure of eukaryotic protein complexes

In organisms, the interaction between protein monomers is crucial

Both RoseTTAFold and AlphaFold can accurately predict protein structure based on the gene sequence that produces the protein, but their strategies are different

It is the different nature of the two sets of tools that give the research team the opportunity to combine the advantages of the two to screen for possible interacting proteins and predict their structure within the scope of the whole proteome

▲The three-dimensional structure of the protein complex predicted by the latest research (Image source: Professor Cong Qian/University of Texas Southwestern Medical Center)

Professor Qian Cong, who currently works at the University of Texas Southwestern Medical Center, said that when she was a postdoctoral fellow in Professor Baker’s laboratory, her research direction was to use the co-evolution between different proteins to predict the parts of the proteome that might interact.

In the latest research, the international team led by Professor Baker and Professor Cong Qian expanded the two AI tools originally used to predict the structure of protein monomers to predict the structure of protein complexes and predict within the scope of the proteome.

The research object selected by the research team is a common eukaryotic model organism-yeast

Subsequently, the researchers used co-evolution between different protein residues to screen 8.

From these three-dimensional structures predicted for the first time, further research has found protein complexes related to a series of functions.

▲A series of three-dimensional structures related to DNA transcription, translation and repair predicted by the latest research (picture source: reference [1])

After achieving fruitful results from yeast, another goal of the research team is naturally to apply this set of tools to human proteins

Note: The original text has been deleted

Reference materials:

[1] Ian R.

[2] Deep-learning in protein-protein interactions identifies complexes that will advance our understanding of cellular processes.