Dr. Tawbi and his colleagues examined the tumor tissues of 21 patients with metastatic melanoma, some of whom responded to chemotherapy and some who did not. Once the cases were divided, the researchers used a mathematical tool called Neural Network Analysis to survey over 25,000 genes and the regulators that turn the genes on and off to see if they could identify ones that could distinguish responders from nonresponders.

"Cancer cells contain massive amounts of information that, if analyzed appropriately, may inform us how to kill them," said Dr. Tawbi. "They contain thousands of genes, and every gene has a switch that turns it on or off. Neural Network Analysis, which utilizes pattern recognition algorithms, helped us identify a signature of eight genes and their switches that predict a patient's likelihood of responding to treatment for metastatic melanoma."