To my mind (as a cancer biologist), the big caution here is in the mouse xenograft model. While it is good that they tested some different cell lines (in a dish), it is much easier to cure a xenograft tumor than it is to cure a sporadic tumor in a human. This is probably because culturing a cell line in the lab for a while tends to reduce the genetic heterogeneity among the cells, and so reduces the chance that there will be a resistant mutant among those cells. Clinical experience shows that there is very often a resistant mutant in the sporadic tumor of a human patient (particularly if the cancer has managed to spread, i.e., metastasized). In any case, they only tested one cell line in the mice. So it is time to try more.

Second off, the cancer stem cell hypothesis is highly controversial, and has been mostly demolished in breast cancer (the type of cancer in this experiment), where Kornelia Polyak has shown that what people are calling cancer stem cells, are probably just different clones / cell types in the tumor. That being said, if we can come up with drug combinations that can handle all the different cell types in a tumor, we'll be doing a lot better than present.

It certainly is promising that the combined therapy showed no relapse over ~60 days after the end of therapy, but I'd like to see the results taking out further to see if the mice would relapse later. I'd also like to see it demonstrated on other cell lines and more realistic models of cancer. There is a long history of drugs looking good in simple "pre-clinical" models, like mouse xenograft models, and then failing in clinical trials.

Also note that it usually takes decades between initial discovery and changing clinical practice, so don't hold your breath, but do cross your fingers.

Actually, the few times I know of that a good data set was put up on the web, it generated a lot of research and progress. I'm thinking of Pat Brown putting up some of the first data on gene expression arrays. Probably hundreds of people worked on that data - everything from statistical methods, to reverse engineering the gene network. It was great. This is probably most valuable when the data is from a new type of experiment that is likely to be widely used.

I hope to do something similar but there is a big problem for geneticists like me. If you post your volunteer's genetic data on the web, there is no way to anonymize it. It would be a simple thing for a medical insurance company to take a cheek swab, run the genetics and then match it against all public datasets to see if an applicant has a known disease. I know of patients that have lost their medical insurance because their insurer found out that they were participating in a research study, and inferred (incorrectly) that the patient had a disease.

I'm a cancer researcher and I agree. Though I'm in it more for the good of society and because it is an engaging problem. I would jump at the chance to cure cancer even if it put my institution out of business and I didn't get the recognition. The reality (of this fantasy) is that most institutions and researchers could easily move on to other diseases/problems. We do it all the time.

In addition, there is BIG money to be made from a drug that cures cancer. Even the ones that cure a small percent of cancer can make Gigabucks these days. This is why big pharma really does try to find new cancer drugs.