PROVIDENCE, R.I. -- Compared with a long-used linear model, a correlation-based statistical method is a more reliable way to map complex gene interactions and pinpoint genes that may be potential cancer treatment targets, according to new Brown University research.
The research is important because it describes a promising new tool for tracing human gene connections, a task critical for understanding and treating cancer and other diseases. Results appeared this week in the online edition of the Proceedings of the National Academy of Sciences.
"Genes influence one another in many intricate ways," said Leon Cooper, professor of physics and neuroscience and director of the Institute for Brain and Neural Systems at Brown. "What we need is a map, or network, of these links. What we've identified in this project is a more effective method for making this map."
The research team – which included scientists from the fields of biology, physics, statistics and computer science at Brown, Università di Bologna in Italy and Tel Aviv University in Israel – set out to answer a question. When a deadly "oncoprotein" is switched on, what chain reaction of gene activity does it set off?
The protein, c-Myc, causes cells to multiply. If the protein is produced unchecked, it can cause breast, colon and other types of cancer. C-Myc contributes to more than 70,000 deaths in the United States each year.
Once the c-Myc switch is thrown, thousands of other genes start pumping out proteins or switching on other genes, which activates still more genes. One way to study this web of connections would be to set off the chain reaction and study it over time. To make that happen, Brown researchers came up with a clever experiment.
John Sedivy, a long-time c-Myc researcher and the director of Brown's Center for Genomics and Proteomics, developed rat cells that lacked the c-Myc gene. These cells were further modified to make a form of the c-Myc protein, which could be switched on or off by the hormone treatment tamoxifen.
One batch of cells was treated with tamoxifen, then harvested one, two, four, eight and 16 hours later. Another batch of cells didn't get the drug but were harvested during the same time frame.
Analysis of gene activity generated in the experiments revealed 1,191 possible players in the c-Myc gene network. A statistical team, led by Gastone Castellani, an associate research professor with the Institute for Brain and Neural Systems and a professor at the Università di Bologna, tested two methods to try to model this network.
One was the linear Markov model, a decades-old tool used to crunch everything from sports statistics to language production. The other was a correlation method based on network theory, which has been used to explain complex systems such as power grids and neural networks.
After applying both statistical methods to the experimental data, the team found that the correlation method was a more effective analytical tool. The method was sensitive enough to capture gene network changes after tamoxifen treatment, producing a list of 130 genes significantly altered by c-Myc activation. This method was also reliable. When researchers reshuffled the data time points, those network changes disappeared.
In contrast, the gene network constructed by the linear Markov model appeared to be insensitive to the effects of tamoxifen. Even when researchers shuffled the data time points, the network appeared largely unchanged.
"Network theory has been hugely informative in analyzing the genomes of simple species such as yeast," Sedivy said. "Here, the theory is applied to a much more complex system: humans. The overall concept – the time series experiments and the combination of statistics and network theory – is quite novel. This should be an important new approach to studying gene expression."
Source: Eurekalert & othersLast reviewed: By John M. Grohol, Psy.D. on 21 Feb 2009
Published on PsychCentral.com. All rights reserved.
It's not having been in the dark house, but having left it, that counts.
-- Theodore Roosevelt