NEW YORK, May 10 – In a paper in Science last week, researchers at the Institute for Systems Biology in Seattle showed they could apply their new hypothesis-free style of scientific research to a real-life example: the galactose utilization pathway in the yeast Saccharomyces cerevisiae.
But researchers say this is only the beginning. Demonstrating the technique in yeast is an important proof of principle, but the goal is to study higher eukaryotes such as humans, said Trey Ideker, the first author of the paper.
“Ultimately the technique will have lots of medical applications, such as elucidating disease pathways in humans, and coming up with new drug targets based on that knowledge,” said Ideker.
Before this can happen, however, new technologies--both biological and computational--will have to come online. Compared to the galactose utilization pathway in yeast, a relatively well-characterized biological system, applying the systems biology approach to humans will require the ability to generate and handle exponentially larger and more complex amounts of data.
“While clearly an impressive demonstration of the power of their approach and technology, there is considerable room for continued development of technologies” to make gene expression studies more accurate and precise, said John Wiktorowicz, the director for proteomics at Lynx Therapeutics, in Hayward, Calif.
Ideker and his co-authors, who include Lee Hood, the director of the Institute for Systems Biology, have at least spelled out the recipe. Rather than study the minutiae of one particular interaction or subset of interactions in a pathway, the authors describe a method for studying the entire pathway, by combining model-based predictions with biological experiments.
In the first step, as described in the paper, the researchers built a model of the galactose utilization pathway in yeast using information on molecular interactions found in the publicly-available scientific literature.
Meanwhile, in the lab, the researchers genetically engineered yeast cells to over- or under-express certain proteins, or varied the cells’ environment, and measured the cells’ response using DNA microarrays and tandem mass spectrometry with the help of a technique called ICAT, a method for detecting proteins found only in low quantities in the cell.
By comparing the model’s predictions with the experimental responses to outside stimuli, Hood’s team could observe where the model predicted the correct response, and where the model did not. In this way, the researchers could gather information about the pathway, while also pinpointing areas for more focused study.
“This is the first time it’s been suggested that you could do these kinds of analyses in a global way,” said David Goodlett, a co-author of the paper and a researcher at the institute.
One of the advantages of this approach, said Goodlett, is that it isn’t necessary for a researcher to formulate a hypothesis before conducting the experiment. Instead, learning where the model fails allows the researcher to propose new hypotheses about how the pathway works, and then go back and test them.
The idea, said Benno Schwikowski, a researcher at the Institute of Systems Biology who contributed to the study, is “to open up hypothesis-driven science to discovery-based methods” by combining powerful computational methods with high-throughput experimental techniques.
To apply this method to pathways in human cells, researchers must first find more efficient methods of mining the public databases for relevant biological information, said Ideker. “We can’t get it out of the literature fast enough,” he said.
In addition, new algorithms will be required for associating the observable changes in gene expression with the underlying molecular interactions that cause them.
On the biological front, regulating gene expression in human cells is more difficult, not only because the number of different types of human cells is so varied, but also because at a fundamental level, yeast cells just have just one copy of each gene compared to two in humans, simplifying the gene’s suppression.
Creating a single chip that contains all the genes in the human genome would certainly help to overcome these obstacles, said Ideker, as would new techniques for regulating genes using antisense RNA.“All these methods are coming,” said Ideker, “it’s just a matter of time.”