Steinman’s last experiment would be, in many ways, the culmination of a new trend in cancer research: designing custom treatments for each patient. When he got sick, Steinman knew that the five-year survival rate for his kind of tumor was, at most, 1 in 10, even at Sloan-Kettering, one of the best oncology centers in the world. Typically, patients live six months. But he also knew that his chances might not be as bad as they looked. The means and medians of his disease were drawn from populations and so did not reflect the fact that every tumor is unique. Even tumors that look the same — cancers starting from a common organ, or a common kind of cell — may behave in different ways: some shrink and some expand; some succumb to chemotherapy. Now doctors can scan each tumor for clues about its DNA and use those clues to determine its strengths and weaknesses. Steinman could have his case described right down to the letters of its genome, in hopes of figuring out which therapies might work best for him.

This “personalized” approach to treating cancer, which subdivides the classic types according to distortions in their genes, has been growing at a rapid pace. In the past few years, laboratories financed by the government have set out to build a comprehensive atlas of the cancer genome — to collect 500 tumors from each of 25 kinds of the disease and then to analyze their DNA and RNA at a cost of more than $100 million a year. The advent of inexpensive genome sequencing has produced a gold rush in the commercial sector, too, with the promise that anyone’s tumor can be sliced and processed and analyzed, until its genetic fingerprint is decoded.

“It was thought a while ago that cancer would be too complex for us to really get our hands around it,” says Raju Kucherlapati, one of the principal investigators on the Cancer Genome Atlas and a professor of genetics at Harvard Medical School. But current research showed that “the total number of major biochemical pathways that are altered is not limitless.” If that’s true, then doctors might use these genomic data to improve their patients’ odds. Instead of applying a one-size-fits-all approach to treatment, they could select a mix of therapies from a standard arsenal, choosing only those that matched the features of a patient’s tumor. “I would venture to say that within the next 10 years, we could see a very significant revolution in the way that we think about and treat cancer,” Kucherlapati says.