I've only looked at the abstract (damn publishers), but superficially this seems to be a great study. I'd just like to underscore how amazing their samples are.

I've only looked at the abstract (damn publishers), but superficially this seems to be a great study. I'd just like to underscore how amazing their samples are. The mean survival time of their patients, 2 years, is also surprising. (Survival bias?) Getting at metastases is hard. In general it simply isn't medically necessary to biopsy them. Without medical necessity it's unjustifiable to put a patient through that risk and discomfort. Why are metastases important? Mostly because they're presumably cleaner than primary tumors. In a primary tumor you start out some founder cell that has lost the ability to undergo programmed cell death. In general, DNA repair mechanisms are also awry. The cell divides producing daughter cells, the mutations add up and so on. The tumor becomes heterogeneous. The daughter cells diverge from one another. If you sample from the primary tumor you get plenty of mutations related to cancer, and plenty of mutations unrelated to cancer, simply because the genome is a big place. It is no fun to work with this data. OK. Metastases. Some immortalized cell has lost basal lamina adhesion and migrated to another site in the body. This is a selective bottleneck -- all of those randomly mutated cells that didn't lose adhesion are stuck in the primary. The secondary should therefore be enriched for mutations in oncogenes and tumor suppressors. This is why this data is so important and could give some great insights into oncogenesis. Heck, here's an experiment I would love to run: * Collect biopsies from primary tumors in non-metastatic cases. * Collect biopsies from metastases in another patient population. * Look for mutations and changes in genome architecture that are unique/overrepresented in the metastatic cases.