Science Editor

MONTREAL – At the American Society of Human Genetics (ASHG) annual meeting, the contrast between the imagined future and the present is sometimes jarring.

On the one hand, sequencing technology continues to improve at a rate that surprises even experts, one panel participant – though acknowledging he was "sort of in hype mode" – imagined a future where genome sequencing would come free with the newest version of the iPhone.

But at a panel on Cancer Genomics, which provided updates from the Cancer Genome Atlas and the International Cancer Genome Consortium, it was also obvious that sequencing has a way to go – and that for getting a high-quality cancer genome, the sequencing itself is only one challenge.

With two talks focusing at least in part on pancreatic cancer, some of the other issues were visible in particularly stark relief in that tumor type. On the panel, John McPherson of the Ontario Institute for Cancer Research described the challenges for sample collection that need to be overcome before his team can even get to sequencing.

Pancreatic cancer accounts for 2 percent of all cancers. But with no good early detection methods, more than half of all cancers have metastasized by the time they are diagnosed and the overall five-year survival rate is an abysmal 2 percent. With four years survival, Nobelist Ralph Steinman had beaten the odds by a considerable amount. (Steve Jobs, who was diagnosed with his cancer in 2003, had a pancreatic neuroendocrine tumor, a subtype of pancreatic cancer that starts with islet cells and has a less-dismal outlook, at least in the short term.)

The low five-year survival rate has several consequences for scientists' ability to get tumor samples. For one thing, McPherson explained, starting out, the International Cancer Genome Consortium required that a very specific consent form be signed – and though his team had a bank of tumor samples from pancreatic tumors, most of the corresponding patients had long-since passed away: "So we had to start from scratch."

His team is planning to sequence 350 tumors and has collected 113 samples to date. Those samples, though, are not the pure tumor cells that would make sequencing them easier. Pancreatic cancer grows in a way that is extremely intertwined with the stroma, or microenvironment.

The fact that tumor and stroma are so intertwined means that only 15 percent of pancreatic tumors can be surgically removed – which leads to another question: Are the sequences of those 15 percent in fact a good stand-in for the other 85 percent? Or are they so genetically different that sequencing them to gain insight into pancreatic tumors more generally amounts to looking for your wallet under a streetlamp?

Theoretically – and in many tumor types, in practice as well – biopsies can fill in some of the genetic blanks for tumors that cannot be surgically removed altogether. But to biopsy a pancreatic tumor is not, to say the least, an outpatient procedure. The way to the pancreas is through the liver, an invasive procedure which carries a risk of bleeding. It's not a procedure for a patient with a life expectancy of three to six months, no matter how much he or she cares about helping future victims of the disease.

One way to get tumor samples is through autopsies within two hours or so after death. McPherson said his team has some such samples from inoperable tumors, as well as metastases, and is working on sequencing them to compare them to operable pancreatic tumors.

The same love of stroma that makes so many pancreatic tumors inoperable, though, continues to cause trouble when the samples are being sequenced. The typical fraction of tumor cells in a tumor varies from 20 percent to 80 percent – "I've seen some with 5 percent," McPherson said. This dependence on stroma, in addition to making the cells hard to culture, means that scientists are often sequencing more stroma cells than true tumor cells.

Sequencing stroma cells is not necessarily a bad idea – in fact, McPherson and his team are looking at stroma cells to see whether there is anything in them that can account for the litany of failed pancreatic cancer clinical trials. But the mix of cells makes it more challenging to get a clean genomic picture of either one.

Between that tumor-stroma mix to the typical heterogeneity of tumor cells, Stacy Gabriel, of the National Institutes of Health, had a sobering warning about how long it will take for the future of cheap and accurate sequencing to come to the more challenging cancer genomes.

Providing an update on The Cancer Genome Atlas that included data on pancreatic tumors, Gabriel warned that "we often talk about 30x genome coverage" – that is, sequencing each sample 30 times to correct for sequencing errors and sample impurities – "as if it's going to be enough. And I think that in these tumor samples, it's not going to be enough – we're going to need 60x or 100x coverage. Or more."