A Conversation With
A Conversation about Sequencing Cancer Genomes with Dr. Elaine Mardis
Dr. Elaine Mardis is co-director of The Genome Center at Washington University in St. Louis. As director of technology development, she leads the center’s efforts to explore progress in next-generation sequencing technologies. Dr. Mardis was part of the team that sequenced the genomes of two adults with leukemia, and she is now working on the Pediatric Cancer Genome Project, which is a partnership between Washington University and St. Jude Children’s Research Hospital.
How many tumors does one need to sequence to get results that benefit patients?
I don’t think anybody knows the answer to that question. While I’m hesitant to speculate on the number, I’m very hopeful, based on our own experience, that we’ll be able to achieve some really significant findings in a short period of time. We’ve already found key mutations that are recurrent throughout acute myeloid leukemia, that have prognostic value, and that in the long term may turn out to be drug targets.
What happens after the sequencing?
Our sequencing activities will generate a list of clues, if you will. By clues, I mean key genes that are mutated in these tumors and germline (inherited) variations that likely play a role. We have to then take that set of clues and try to understand the implications of these mutations for the development of cancer. This work requires cell culture models, mouse models—all of those things are components of the research network and infrastructure that exist now at St. Jude.
I think it will be very exciting to work with that aspect of their operations and fill their pipelines, if you will, with all of these clues. This, in turn, will further our understanding of the basic impact of those mutations on cancer biology and the development of the disease. At the end of the day we need to have that infrastructure and expertise in place so that we can gain a better understanding of this disease, which is obviously what this is all about.
Do you have estimates for the costs of the sequencing?
It’s possible to do an estimate right now today and then the number will actually change dramatically over the upcoming months. But I can give you a sense of what the trajectory has been and probably will be.
Back in ‘07 and ’08 when we sequenced our first leukemia patient’s genome, we completed the project for around $1.5 million. We were really doing two genomes, because we were comparing the DNA of tumor cells and normal cells. It was quite an expensive project, but this was the first time that had ever been done. We had to build a lot of the infrastructure and the informatics to analyze the sequence data from the ground up, and that’s now in place. We now have a well-honed and, in many senses, an automated process for analyzing next-generation sequence data.
Our second tumor-normal genome comparison cost about $500,000, so already there was a dramatic drop in the price. That had to do with two things. We didn’t have to spend as much time generating the data, because the ability to sequence on the instruments we were using increased. The other reason was that many of the bioinformatic programs were already in place to analyze the data.
If you jump forward to today, it now takes just two runs of the instrument to generate the information for the tumor sample and another two runs to generate the data for the normal sample. We’re looking at a cost for that tumor-normal pair of about $90,000. So you can see a rather dramatic drop in cost in a reasonably truncated time frame, and the price keeps falling successively. So even though the St. Jude proposal specifies 600 tumor-normal pairs, it’s quite likely that with the same amount of money in this collaboration we’ll be able to do significantly more than that over the 3-year time period just because of that cost reduction.
Can you talk about your group’s ability to manage and make sense of the data?
One of the forward-looking things that the dean of the medical school, Dr. Larry Shapiro, did about 3 years ago was to allow us to build a state-of-the-art data center. We have filled the space with around 6,000 computing cores and over 5 petabytes of data storage capability. So this infrastructure is in place along with bioinformatic programs to do data analysis, which were largely developed for our work on adult cancers.
The St. Jude opportunity builds upon this existing infrastructure and bioinformatic capability. Working closely with our collaborators at St. Jude, we now need to figure out which of the samples and which of the clinical questions in the near term will be most relevant to some of the weightier questions in pediatric cancer we’d like to explore. We will then get those samples up here and start on DNA sequencing.
How will you integrate the genomic and the clinical data?
All of the samples are from patients treated at St. Jude, and the samples come to us with extensive clinical records, so we know how the patients did, how they were treated, what the outcomes were. It is really invaluable to have that kind of rich clinical data available to us. And in many cases we will be dealing directly with the oncologists who acquired these samples and treated these patients. These oncologists, of course, have an intimate knowledge of the disease.
Once we have the data sets in hand and the preliminary analyses completed, this, in my mind, is where the enterprise succeeds or fails. The challenge will be to roll up our sleeves with those pediatric oncologists and begin to discuss what the data might mean—and how to interpret these results in the context of that clinical information. This project really involves building a strong collaborative relationship with those key players at St. Jude. We already have that with Dr. James Downing and some of his leukemia colleagues, and we’ll continue to build that out as we look at the other major tumor types that occur in children.
What might you learn about normal genomes from this project?
For starters, I don’t think that we know enough about normal genomes. In some ways, the beauty of sequencing the normal genome along with sequencing a tumor is that, over time, we accrue so much information about normal genomes and the variation that lies therein. In fact, we have an active effort in our existing cancer-sequencing program to build a database of normal genome variation. To look at the big picture, cancer, regardless of whether it is pediatric or adult, is an interplay between the mutations you are born with—your germline (inherited) mutations—and whatever happens downstream of that—your somatic (naturally occurring) mutations.
Now, in pediatric cancer this is a particularly interesting question. Obviously, adult cancers make more sense in that the disease happens to people who are older. Their bodies have experienced many cell divisions, and there have been many opportunities for mutations to creep in. But for children, that logic does not apply. So I think there is likely to be a story here looking at normal variation. We are very keen to generate lots of data on normal genome variants and see what we can come up with in terms of that interplay between the germline and the somatic mutations.