I was really excited to see several papers on "cancer genomes" appearing just before christmas. There were two in Nature's advance online service mid-December, widely hyped in the press, and then another one in the print issue of 24.12.
However when I read the lung cancer paper I found it quite disappointing. As I understand it, the authors sequenced 25-year-old cell lines from a patient whose smoking history is unknown. While there is a 97 % probability that his cancer was caused by smoking, this also means that there is a 3% risk that the conclusions are just wrong and the mutations relate to a different cause. And for any specific mutation one might be interested in, one simply doesn't know whether it developed in the cancer or in the cell culture, and whether it's a "driver" mutation (i.e. promoting the cancer) or a "passenger" mutation that just hitched a ride. Also, the control sample will have developed somatic mutations as well, so one would have needed several controls to figure out what is a relevant mutation. The authors admit somewhere that the various rates of somatic mutations aren't very well understood yet.
As millions of smokers get lung cancer every year, I really don't understand why the researchers couldn't get a fresh sample linked to a well-documented case study. Insiders tell me that it's tricky to sort out the legal side of doing such studies on a real patient, but then again, that's what the whole thing is for, namely ultimately to be able to analyse a patient's cancer, to find out how it can be treated most efficiently. And if I went through all the trouble of sequencing two human genomes (cancer and control) several times over, I real would want to make sure that the information I get can be interpreted meaningfully.
The second online paper also used cell lines, this time from a melanoma (so at least there is no doubt that the patient has been exposed to sun light, though again, we probably don't know any further details).
The paper in the print edition of 24.12. (vol. 462, p.1005) doesn't look at complete genomes but samples rearrangements of genetic material. This time, at least, there are several "fresh" samples included among the 24 separate breast cancer cases studied. So it can be done, and I think this is the way forward. The other two papers only show that one can analyse the genomes of cell lines - which I think we all knew before. Now we need lots of genomes of lots of real tumours, with real case histories, and then we can start discussing what caused these cancers, and which are the "driver" mutations, and which the "passenger" mutations.
In the newspaper coverage I read, the two online papers based on cell lines were represented as if they were based on actual living patients, so this subtlety clearly got lost in translation.