You're a stressed-out intensive care unit (ICU) doctor, trying desperately to save the lives of a score of patients under your care. Most of them are on ventilators, and you know that a good number of such patients will go on to develop pneumonia, upping their risk of death. You want to treat them, and fast.

But which ones do you treat? Do you treat all of them with antibiotics, which could lead to other complications? Or do you wait until you can grow pneumonia bugs in the lab, rendering test results too late to prevent death?

A new test under development could allow doctors to figure out which patients need treatment, before the pneumonia shows up. It makes use of gene chips, also known as microarrays—technology that has been used for more than a decade in the lab to study genetic variations and how a particular gene is turned on and off among humans, other animals and plants. There are clinical situations in which doctors are already using it. For instance, in 2007 the U.S. Food and Drug Administration approved a microarray created by Amsterdam-based Agendia called MammaPrint, which helps oncologists predict whether a woman's breast cancer might return in five or 10 years.

J. Perren Cobb, director of Washington University School of Medicine's Center for Critical Illness and Health Engineering in St. Louis, however, is the first to use microarrays to map deadly infections before symptoms begin in such an acute setting.

Cobb's team studied patients in an ICU before they had symptoms of ventilator-associated pneumonia, using microarrays to find 85 genes that were active if a patient was going to develop ventilator-associated pneumonia. This gene activity for the ventilator pneumonia could be distinguished from that activity in lab mice with the disease from a different type of bacterium. Cobb thinks that he has discovered a way to predict which patients will come down with pneumonia before the symptoms start. Although he did not use that information to make treatment decisions, he hopes that one day doctors can use this approach to save those individuals from the painful and costly complications brought on by spending weeks hooked up to a ventilator.

Like Cobb, Jeanine Wiener-Kronish, chief of anesthesiology and critical care at Massachusetts General Hospital in Boston, wants to find more precise tools that allow doctors to preemptively treat the right patients, save lives, and save hospitals the $50,000 per case that ventilator-associated pneumonia can cost. It is better than the reactionary approach doctors have taken for the past century, says Wiener-Kronish, adding, "What we do now is just like voodoo."

Despite his results, Cobb says using microarrays to diagnose infection is not currently practical. At less than $1,000 per patient, microarrays cost less than using CT scans in the ICU, but answers take several days—too long to diagnose an infection. Still, Wiener-Kronish says Cobb's research will help alert doctors to the onset of an infection and distinguish different types of infections. "His idea that there's a pattern in the immune response is right on," says Wiener-Kronish.

In the meantime, Cobb's team is using microarrays to detect ventilator-associated pneumonia on a larger group of ICU patients and making plans to use this test in a clinical trial. He one day envisions that just as an electrocardiogram (EKG), the squiggly lines that trace the electrical activity in the heart, can tell a doctor whether you've had a heart attack, a test could record changes in gene activity and tell doctors how your immune system is doing—an immune checkup of sorts.

That kind of widespread use—for example, the U.S. Department of Defense could use the immune checkup to test for the flu in a submarine crew before a long mission—would require faster versions of technology than is currently available.

Then, doctors could use microarrays or a future technology to monitor diseases such as rheumatoid arthritis, because the immune cells that fight infection are often the same ones that cause inflammation. Cobb's team—which is funded entirely by the U.S. National Institutes of Health—is working with a group of doctors that treats such diseases to study the technique's effectiveness. When they used their immune checkup on 20 infants with infections, they found 14 genes that were more active. Those could be used—instead of a spinal tap—to figure out which babies who show up at an emergency room with a fever get immediate antibiotics, and which will not need treatment. What was more, in the study, presented in October 2007 at the American Academy of Pediatrics, five of those 14 genes could be used to distinguish viral infections from bacterial infections, giving doctors in the future more information with which to make quick decisions.

To move the research forward, Cobb, Wiener-Kronish and others will meet in November to plan the first clinical trials of the technique at the ICU bedside. "We hope to focus attention on our patients who need it most," he says, "leveraging the power of the human genome, for their benefit and their family's."