For several years now, large-scale cancer-genome studies have made it increasingly clear that a tumour cell is a genetic disaster area littered with mutations that differ not only from one type of cancer to the next, but from one patient to the next. Pharmaceutical companies have had to accept that Gleevec, a drug that treats a form of leukaemia by targeting a specific gene product, is almost certainly going to be a rare exception in the therapeutic arsenal; most cancers are far too complex to yield to such a magic bullet.

That message was hammered home with new statistical power in three studies released last week (see page 148). Two of the studies, published in Science by a team based at Johns Hopkins Kimmel Cancer Center in Baltimore, Maryland, focused on pancreatic cancer and a type of brain cancer called glioblastoma multiforme — both among the most fatal and intractable tumours known. The third paper, published in Nature by the Cancer Genome Atlas project, also focused on glioblastoma. The studies took a more comprehensive approach than previous large cancer-genomics studies, by simultaneously analysing genetic sequences, copy-number variations, expression arrays and other forms of data. The Johns Hopkins team looked at all the active genes in tumours from a few dozen patients; the Genome Atlas team looked at selected genes in tumours from 206 patients. Taken together, their results show that no single mutated gene lies at the heart of any of these tumours. The pancreatic tumour samples, for example, showed an average of 63 genetic mutations each — with considerable variation from one sample to the next.

That conclusion might make the prospects for new targeted drug therapies for cancer seem hopeless. And yet, the reality may be just the opposite. The richness of the data becoming available in these and other studies allows researchers to cut through the complexity. Genes work together in pathways of reactions to accomplish a particular biological function, such as cell division — and many or most of the mutated genes picked up by these cancer studies are involved in a comparatively small number of pathways. The Johns Hopkins team found that most of the mutations in their pancreatic tumours affected just 12 pathways. The Genome Atlas team found that most of its glioblastomas showed mutations in a set of three pathways. So drugs targeting these pathways might work in more patients than drugs that target only one of a pathway's myriad gene components.

To realize that hope, researchers and funding agencies will need to do many more such studies on many more types of cancer. Just as important is the next step, which is to determine how these mutated pathways contribute to the development of cancer — and how that contribution might be removed. After that comes the task of finding useful biomarkers, chemical signals that will allow therapists to determine which pathways have been affected in each cancer patient, and how that patient will respond to any given therapy.

None of this will be easy. Untangling the immense complexity of cancer will be big science by anyone's definition, requiring a long-term commitment and enormous amounts of data. And yet, that very complexity has begun to give reason for optimism.