Computer-Generated Proteins

Groundbreaking, a milestone, and an impressive technical achievement. This is what some scientists are calling the work of researchers at the UW and Fred Hutchinson Cancer Research Center who have developed a computational method that can accurately predict the structure of a custom-designed protein not found in nature.

The research project was headed by Dr. David Baker, associate professor of biochemistry and investigator at the Howard Hughes Medical Institute; Dr. Brian Kuhlman, formerly a postdoctoral researcher in Baker's lab and now assistant professor of biochemistry at the University of North Carolina at Chapel Hill; Gautam Dantas, a graduate student in Baker's lab; and other collaborators.

The team's research addresses two problems: predicting the sequence of amino acids when given a protein structure and predicting the protein structure when given the amino acid sequence.

Though previous research had shown that computer programs could predict an amino acid sequence for a naturally occurring protein structure or a slightly modified one, no computer model existed for arbitrary protein structures that have not been observed in nature.

Baker's group designed just such a model, which predicted with great accuracy the amino acid sequence that would fold into the particular protein structure. The researchers tested the program by using the predicted amino acid sequence to synthesize a protein, and then comparing its structure to the structure of the theoretical protein.

The protein and design calculations are carried out using a computer program called Rosetta. To predict a structure from an amino acid sequence, the program searches for the lowest energy 3-D structure for that sequence. To design a sequence that will fold up to a desired structure, the program searches through all the different shape combinations of the 20 amino acids for the lowest energy sequence for the desired structure. Both the physical model and the search algorithms are continually being improved based on feedback from the prediction and design tests.

The biggest challenge of computational protein design is the creation of novel proteins with arbitrarily chosen 3-D structures. This challenge has been realized with the development of a new protein called Top7 with a novel sequence and topology. The X-ray crystal structure of Top7 is strikingly similar to the design model. Top7 was designed and synthesized using protein design-sequence algorithms and the group's structure-prediction program, Rosetta. It is the first globular protein with a new fold that was designed from scratch and validated by its X-ray crystal structure.

The significance of this research is enormous. Designing arbitrary protein structures can help researchers better understand naturally occurring proteins and how they evolved, as well as provide potential tools for drug treatments that may not be possible with proteins found in nature. Custom-designed proteins could also possibly be used as catalysts in reactions and act as motors for moving things in new ways at the cellular level.

The results in Baker's lab have opened up the possibility of custom-designing potentially useful proteins that nature hasn't gotten around to devising yet.

"Our ability to design a new protein from scratch shows that we are not far from a fundamental understanding of the energetics of intramolecular interactions," Baker said. "On a more practical side, we can now think about designing new proteins for a wide range of applications."

At international blind tests of structure prediction methods carried out over the past six years, CASP3, CASP4 and CASP5 (Third, fourth and fifth Community Wide Experiment on the Critical Assessment of Techniques for Protein Structure) at Asilomar in Pacific Grove, Calif., Baker's team has been consistently ranked the best in the world at interpreting amino acid sequences and predicting their 3-D structures.