Another layer of complexity: short lengths of RNA could provide new form of genetic control.

Most of the human genome is so-called junk DNA, which contains no code for proteins and was long thought to be useless. Now, researchers have found that a relatively large portion of this genetic material could help regulate the activity of nearby genes.

"This points to a new layer of control and complexity of the genome," says Aarron Willingham, a member of the team that conducted the study at Affymetrix, a Santa Clara, Calif., company that sells DNA-analysis tools.

The new research could explain a puzzling 2002 discovery by Affymetrix researchers. They found that human cells transcribe large regions of junk DNA into strands of RNA. For true genes, this RNA would serve as a template for making proteins, but for junk DNA, producing RNA seemed like a waste of a cell's energy.

The new study shows that the extra RNA may have a function after all. The long strands of RNA serve as a raw material from which the cells make snippets of RNA containing fewer than 200 "letters" of the genetic code. The researchers also found an apparent link between those short strands, called sRNAS, and the activity of certain genes, suggesting that the sRNAS might influence a cell's behavior.

After examining thousands of active and inactive genes in cultured human-liver and -cervix cells, Willingham and his colleagues found that about 44 percent of the active genes had sRNA-encoding regions nearby, while only 20 percent of inactive genes had such neighbors. While the mechanism tying sRNAS to the genes' activities remains unknown, the researchers say that the correlation is striking.

"I'm really impressed," comments Katalin Fejes Toth, who studies RNA at Cold Spring Harbor (N.Y.) Laboratory. "The fact that they looked at thousands of genes makes it more believable."

"It does strongly point to a regulatory role" for sRNAS, agrees Thomas Gingeras of Affymetrix, who led both the 2002 study and the new one, reported online and in an upcoming Science.

An alternative explanation is that DNA coding for an sRNA simply gets transcribed by accident along with the nearby gene, Gingeras says. This scenario could give the appearance of a connection without the sRNA serving any function. But the scientists compared the human and mouse genomes and found that the genetic code for SRNAs hasn't changed much over millions of years of evolution, suggesting that the sRNAS play an important role in cells.

If so, this additional of layer of control could help explain how animals, such as people, with more-complex bodies can have a similar number of genes as more-primitive animals, such as roundworms, Gingeras says.

Another kind of RNA, known as microRNA, also plays a role in regulating gene activity. But while microRNAs work by silencing genes, sRNAs correlate with increased gene activity and probably work by a different mechanism, Gingeras says.

The researchers estimate that 1.3 percent of the human genome codes for sRNAs. In comparison, genes make up only about 1 percent of the genome.