- News Home
6 March 2014 1:04 pm ,
Vol. 343 ,
Magdalena Koziol, a former postdoc at Yale University, was the victim of scientific sabotage. Now, she is suing the...
Antiretroviral drugs can protect people from becoming infected by HIV. But so-called pre-exposure prophylaxis, or PrEP...
Two studies show that eating a diet low in protein and high in carbohydrates is linked to a longer, healthier life, and...
Considered an icon of conservation science, researchers at World Wildlife Fund (WWF) headquarters in Washington, D.C.,...
The new atlas, which shows the distribution of important trace metals and other substances, is the first product of...
Early in April, the first of a fleet of environmental monitoring satellites will lift off from Europe's spaceport in...
Since 2000, U.S. government health research agencies have spent almost $1 billion on an effort to churn out thousands...
- 6 March 2014 1:04 pm , Vol. 343 , #6175
- About Us
Calling Cancers With Gene Chips
4 June 2001 7:00 pm
To choose the best therapy for a spreading cancer, doctors need to know which tissue spawned it. Now, researchers have trained a computer program to distinguish the genetic fingerprints of four seemingly similar, deadly tumors. The method could lead to faster diagnosis and more appropriate treatment for a variety of cancers.
Pathologists often pinpoint a cancer's origin by studying the cells' size, shape, and color under a microscope. That's a tricky task for cancer cells that look alike, such as a group of four rare childhood cancers collectively called small round blue cell tumors. In these cases, pathologists often rely on several stains that bind diagnostic proteins, which can delay diagnosis. Betting that DNA fingerprinting would be faster and more accurate, molecular biologist Paul Meltzer and pediatric oncologist Javed Khan of the National Human Genome Research Institute (NHGRI) in Bethesda, Maryland, and their colleagues looked for gene activity in the four cancers--neuroblastoma, rhabdomyosarcoma, non-Hodgkin lymphoma, and Ewing family of tumors.
The team surveyed 6567 genes at once using DNA microarrays. On these so-called gene chips, the activity of each gene is represented by the intensity of a colored dye. To determine the pattern of gene expression, the researchers then ran the color intensity data through a so-called artificial neural network, a program modeled on the behavior of neurons in the brain that can learn by trial and error. "You're getting a complete picture of what's going on with this cancer," Meltzer says.
After the program learned the features of each cancer type, it correctly identified all 20 test samples. It also correctly rejected five samples of other types of cancers as well as healthy cells, the team reports in the June issue of Nature Medicine.
"I think it's very encouraging," says cell biologist Lee Hartwell of the Fred Hutchinson Cancer Research Center in Seattle. Although the sample size is small, he cautions, the work shows clearly that cancers from the same cell and tissue type are similar enough to distinguish them from other cancers, despite genetic differences between people, he says. "If that continues to hold up, these microarrays are going to be very, very important for diagnosis," he says.