This article is from the archives of the UB Reporter.

UB researchers win top spots in global "protein structure Olympics"

Published: August 4, 2005

Contributing Editor

For any institution that is home to even a single winner in the international "protein structure Olympics," winning a top spot means automatic bragging rights.



But until the most recent competition, it was unprecedented for three research teams from the same institution to make the most accurate blind predictions of an unknown protein structure.

In the Sixth Community Wide Experiment on the Critical Assessment of Techniques for Protein Structure Prediction (CASP), three of the 17 winning predictor teams were from UB, and two of them are affiliated with UB's New York State Center of Excellence in Bioinformatics and Life Sciences. A total of 150 predictor teams from universities and scientific institutes around the world participated in the competition.



Scientific papers by the winning teams will be published in a forthcoming special issue of the journal Proteins.

The ultimate purpose of CASP is to find the very best computational methods to elucidate three-dimensional protein structures.

"This really puts UB and its New York State Center of Excellence in Bioinformatics and Life Sciences on the map," said Daniel Fischer, associate professor in the Department of Computer Science and Engineering, School of Engineering and Applied Sciences, who is affiliated with the center and was one of the winners in the competition.



The other winning UB teams were Jeffrey Skolnick, professor of structural biology, and Yang Zhang, assistant professor, both members of the Data Intensive Analytical Bioinformatics Core Group in the Center of Excellence, and Yaoqi Zhou, associate professor, and Hongyi Zhou, research assistant professor, both in the Department of Physiology and Biophysics in the School of Medicine and Biomedical Sciences.

In the biennial competition, several dozen protein amino-acid sequences are provided to bioinformatics scientists who compete with one another—and with automated computer servers using algorithms written by humans—to complete the three-dimensional structures of the same mystery protein.

Skolnick's predictions won in the contest among humans, while Fischer and the team of Yaoqi Zhou and Hongyi Zhou each won for the predictions of their separate computer servers.

"This has established that Buffalo is on the frontlines of protein-structure prediction," Fischer noted.

Predicting structures has been a top priority since the sequencing of the human genome in 2003. The National Institutes of Health has allocated millions of dollars to solve between 10,000 and 20,000 structures experimentally during the next 10 years.

At stake is a far greater understanding of proteins, how they interact and cause disease, and how new drugs might be developed rationally to fight them.

"Because there isn't enough money or time or people to complete an experimental structure for each protein in the world, the idea is to have representative sets of experimental structures so that those for which we don't have structures can be generated computationally," explained Fischer.

In 2000, Fischer was responsible for introducing into CASP the participation of fully automated computer servers. Since then, he has been the organizer of the parallel competition, CAFASP—Critical Assessment of Fully Automated Structure Prediction.

"I wanted to find out how capable automatic programs are in making blind predictions without any human intervention," he said. "The beauty of this is to find which humans are better, which programs are better and what the difference is between humans and programs. So now we have humans versus humans, servers versus humans and servers versus servers, all working to determine the structure of the same proteins. The availability of fully automated programs allows thousands of non-expert biologists to be able to predict the structures of their proteins of interest."

He noted that while the servers are getting better all the time, five to 10 human experts still perform better than the best automated programs.

"But the gap is likely to narrow in the future," said Fischer. "I would not be surprised to see the programs win in the next 'Olympics,' similar to what happened in the famous chess match between Kasparov and Big Blue."