Building IT to Solve The Hardest Puzzle: Human DNABy Larry Dignan | Posted 2005-12-06 Email Print
Celera Genomics' president needs high-powered data filtering to commercialize the human genome.
Celera Genomics, a Rockville, Md.-based biotechnology company, couldn't have cracked the human genome without information technology.
It's a point Kathy Ordoñezpresident of Celera Genomics, which develops drugs to treat psoriasis, cancer and asthma, and its sister company, Celera Diagnostics, which creates genetic testing toolsknows all too well. Celera needs its systems to crunch algorithms and sort through millions of genetic possibilities to find relationships between genesfor, say, taste and smelland a condition like heart disease.
Ordoñez left pharmaceutical giant Hoffmann-LaRoche in 2000 to become president of Celera Diagnostics, based in Alameda, Calif. In 2002, she took the reins of Celera Genomics from founder J. Craig Venter, who pushed Celera to use technology to decipher the human genome. Ordoñez's mandate: commercialize the company's findings.
Celera Genomics and Diagnostics reported a collective operating loss of $140 million on revenue of $66 million for the year ended June 30.
Baseline business editor Larry Dignan recently sat down with Ordoñez to discuss how Celera uses technology to explore the human condition.
How important are information systems to commercializing Celera's discoveries?
The work we do both at Celera Genomics and at Celera Diagnostics is very technical. In both groups, we have very large-scale ongoing discovery programs. At Celera Genomics in Rockville, we have been studying cell-surface antigens [substances that spur the body to create antibodies] in cancer, and what's involved is isolating and purifying those samples, and then looking at patterns. We use that pattern information alongside our understanding of the human genome to figure out which protein resulted in those patterns.
On top of that, we've got a substantial bioinformatics capability categorizing different commercially and publicly available databases and linking them with our own in-house knowledge of genes, so we can pull out all the literature references about that protein and any experiments done internally, and link them to knowledge from other work ongoing anywhere in the company. That aspect of bioinformatics, where you link things to come to new conclusions and drive new experiments, has been very important to us.
Did you build your databases and applications internally, or buy off the shelf?
There are databases of information that are publicly available [such as the National Human Genome Research Institute] that augment our own database. Celera Genomics created the first commercially available database of information about the human genome, and we have constantly updated that and added information from the public. That's been a very important part of what we've done, as well as linking that information with public and commercially available databases.
What role does technology play on the diagnostics side of the business?
At Celera Diagnostics in Alameda, we have been conducting very large genetic disease-association studies. And here we get hundreds of thousands of samples from people with a given condition, let's say cardiovascular disease, and controlled samples, then similar populationsage, gender, etc.of people who did not develop cardiovascular disease. And we've developed assays [tests for a substance] that allow us to interrogate for a one-letter change in the DNA code.
What needs to happen technology-wise to sort through genes?
There are a number of statistical approaches that have evolved. We have developed our own algorithms that process the data [from experiments] every night. That information is automatically downloaded and built into the data collection to make sure we've got the right gender of the right person in a given reaction.
That information is analyzed in terms of how the genotyping calls [changes in the genetic code] were made per reaction, and then reported to algorithms that zero in on where these differences could be. For example, people with cardiovascular disease may have a predisposition for Alzheimer's and other conditions; some have blue eyes and others brown eyes. So there are many differences among these people, and you have to get through that noise and focus on the parameter you're studying.
So, technology to you is about producing the algorithms and databases.
Primarily, yes. But in Alameda, we conduct enormous experiments looking at genetic variation and how it relates to disease. And [the results] have been fully automated and bar-coded so that as the various reactions are assembled to conduct the experiment, we have a proprietary system that coordinates the whole thing.
The people who work in the lab have handheld devices that they can use together with the bar coding on the reaction to keep track of everything. The system is looking for a certain statistical amount of variation that allows us to tell that an individual reaction tray appears to have been processed properly and we can trust the data.
Would your company exist without information technology?
I suppose Celera could exist, but we could not do what we do without information technology. It's enabling for us. We could not have sequenced the human genome. Imagine that.
Given how technology enables your business, how do budget for it?
There are certain aspects of our I.T. spending that I probably look at like any other person you talk with, and that relates to spending for [enterprise planning software] and managing our desktop computers and other things required in any company. I look at that in the traditional sense as percentage of revenue per employee and those types of things.
Within Celera Diagnostics, for example, we manufacture molecular diagnostic products, and one of the projects that we are implementing this year is an automated document system. And that's a pretty traditional I.T. function; where there are commercial sources of these document control systems, we'll make a selection of one, and implement it.
But in the I.T. work that supports our discovery effort, you have to look at that in terms of what's required. In that case, you're making a relative decision of how to spend an R&D dollar. Do you buy new equipment to process algorithms more quickly, or do more genetic experiments? Those are trade-offs.
What needs to happen to get targeted medicine for a genetic condition?
Well, there are a couple of examples alreadyfor instance, diagnostic testing for breast cancer patients who would most benefit from Herceptin. That's an example today. There are some other examples, but we're all waiting for many more and for the technology to move beyond cancer into other diseases.
And it's unfolding now. At Celera Diagnostics and Celera Genomics, we've been conducting very large studies allowing us to substantially increase our understanding of biology. And with that, we are developing new diagnostics, and one of those is in the cardiovascular area. We've identified roughly 20 [genetic] markers that are highly associated or correlated with increased risk for a heart attack. And those markers are now being evaluated in a large population study. We will select the ones that are most useful in predicting risk for a heart attack, and assemble those into what we call a genetic risk score. This test will give each of us a relative indication of our genetic risk for a heart attack.
This is important because today, your doctor estimates your genetic risk by asking you about heart disease in your family. And that's not a very good assessment. First of all, 80% of us do not know what our grandparents died from, and secondly, you don't know which parent is represented in your genes. So you could have a lot of cardiovascular disease on your father's side of the family, but you don't know whether you got that risk or not.
How do you balance the art with the science when filtering out genetic noise?
There's a lot that is not done by the tools and the computers. And that involves, first of all, deciding which problem to go afterdo you study cardiovascular disease first, or Alzheimer's? So, in that particular example, the way we set our priorities was to try to identify the greatest unmet medical need that could potentially be addressed by the technologies that we could bring to bear, and then see whether we could obtain the necessary samples.
That requires human thinking in trying to determine what the priority is. We have to look at different methods that are available now for diagnosis, prediction and monitoring of disease and figure out where we should focus because of unmet needs. For example, diagnosis of breast cancer is imperfect today with mammograms, but at least there is an approach. Whereas the diagnosis of pancreatic cancer is difficult, so that would be a higher priority for us than breast cancer. Algorithms allow us to pull together information, but there's a tremendous amount of creativity needed to connect the dots biologically.
So, humans have the curiosity to pursue these associations, and technology filters the variables.
Right. Take the taste gene. Through our various databases, we should be able to find all papers published about that gene and whether it's been linked to something else. And any patents on that gene. We can very quickly scan the known literature about that gene, and then use that information to develop a hypothesis about how it could be involved and then potentially test that.
So, the human component is probably the most important, but you couldn't begin to do experimentation at this scale without the discovery tools and the technology.