This exercise yielded a set of 48 genes that markedly change their activity depending on their sensitivity to MNNG. When Samson's team used this group of genes to predict the sensitivity of the remaining 16 cell lines, the algorithm anticipated whether or not their growth would be inhibited by the drug with an accuracy of 94 percent.

"This is incredibly encouraging for a clinical application," says Samson, who hopes that her approach will quickly be extended to map the relationships between gene activity and sensitivity to other drugs.

Joanna Peak of Cancer Research UK says that the research highlights "that cancer treatment should no longer follow a 'one size fits all approach'. The accuracy of 48 genes is impressively good, but I would like to see this tested on a much larger panel of hundreds of patients".

The results for MNNG alone have several benefits, explains Samson. The predictive power of the 48 genes, combined with the relative ease and speed at which transcriptional profiles can be compiled for individual patients, make it a very fast and reliable diagnostic tool. But the 48 genes offer further insights into the mechanism of cancer itself that may one day lead to better treatments. For example, when the researchers analyzed the molecular pathways in which the genes participate, they discovered that most of the pathways can be placed in a highly connected network of genes that are already known to affect cancer.

Learning more about the roles of these genes could lead to a better understanding of how the disease can be attacked. Some of the genes could become direct therapeutic targets if reducing or enhancing their activity leads to increased drug sensitivity.

William Phelps of the American Cancer Society says that taking a baseline profile of healthy cells, as the current study does, will be particularly useful for determining a patient's response to therapy before treatment is started.

Samson hopes that the approach will be explored in clinical trials and that tests will go beyond predicting the sensitivity of cancer cells. "I'd love to see this being used to look at normal tissues, not just at tumors," she says. The method could help doctors predict and better protect patients from drug side effects, and it could help medical professionals advise individuals who may be exposed to such compounds in the environment--heavy smokers, for example.