PRINCETON, NJ--Orchid Biocomputer's vision of what the future of genomics research looks like is so clear that it has already made the movie. Construction of Orchid's ultrahigh-throughput single-nucleotide polymorphism (SNP) scoring facility--what CEO Dale Pfost calls the Mega Snipatron--is still underway at a business park here. Yet investors are making virtual visits to a slick computer-generated video version of the high-tech operation, where room after room is outfitted with SNP scoring robots and a shipping and receiving bay stands ready to handle thousands of DNA samples per day. "The next three years are perhaps the most important of the genomics revolution," Pfost, blueprints in hand, tells the camera scanning his construction site. "Orchid will capture the high ground by finding medically important associations to create a whole new range of intellectual property rights and to treat unmet medical needs," Pfost contends, adding, "This facility is going to make that happen."
Indeed, the recent creation of the SNP Consortium, an international group of drug companies that have banded together to create a public SNP database, was confirmation that the pharmaceutical industry is ready to move into the next phase of the genomics revolution. Orchid is charging ahead of them.
In the days before the SNP Consortium was formed, Orchid began generating a barrage of press releases describing its SNP-related activities. In early March, Michael Boyce-Jacino, vice-president of genomics research and development, told a genomics conference in San Francisco that Orchid's new lab would be completed by mid-year and scoring millions of SNPs per day by late 2000. That day Orchid also announced that it had established, with scientists from the University of Cincinnati School of Medicine and the University of Pennsylvania Health Systems, a Clinical Genetics Network by which it will obtain clinical samples for scoring at the Snipatron. Results, Pfost asserted, will enable physicians to "choose the most effective drug for a particular individual." The week following that revelation, Orchid entered a collaborative agreement with Leroy Hood and the University of Washington in Seattle. With Orchid's SNPStream automated genotyping system, the university's Institute for Quantitative Systems Biology would soon begin scoring 30,000 genotypes per day for pharmacogenetic analysis of clinical samples. Orchid also revealed that day that it had secured a $16 million lease-financing package with Oxford Venture Finance. Orchid, which was founded in 1995, had completed a $28 million private placement in 1998 led by OrbiMed Advisors, Invesco Funds, Oxford Bioscience Partners, WPG Farber, SmithKline Beecham, and Motorola. Also in March, Orchid agreed to make Beckman Coulter's Sagian Core System the platform for Orchid's SNPStream technology. Finally, in April, Orchid launched SNPs.com, a web site for SNP information and resources and a venue for scientific discussion about SNPs.
BioInform visited Orchid's offices here recently for a conversation with Pfost about his company's SNP scoring technology and his plans for commercializing Orchid's discoveries. In part one of the interview, Pfost explains the significance of SNPs and describes Orchid's mission. Part two of the interview, in which Pfost talks more about Orchid's technology and future product plans, will appear in the next issue of BioInform.
BioInform: How did Orchid get into SNP genotyping?
Pfost: We had developed a broad platform of technologies and were looking for the applications part. About a year and a half ago we started looking for proprietary biologies and chemistries to put into chips, so that's when we started looking for genomics technologies and specifically pharmacogenomics.
We bought Molecular Tools, a company in Baltimore that had pioneered the ability to look at the differences in DNA from patient to patient to patient, in order to populate our chips with the ability to do simple genotypes.
The first wave of the genomics revolution has been so much physical work in laboratories and industrial settings. The next phase is even more compelling: we are starting to look at the differences in DNA from patient to patient and starting to see how those differences correlate to medically important attributes.
Looking at the next generation of genomics, which we think is the variability of genetics from patient to patient, we start to see that the dynamics and the time frames are quite different. It's all about how you industrialize it and scale it up and how you organize yourself to make data useful. We have the ability to scale dramatically. To look at the differences, instead of using sequencers we're using SNPs and SNP-scoring systems--assay systems that can work with very high throughputs. Instead of looking at every single base in the DNA, we're looking only at those that are the variable ones. We're focusing on only making panels of those. Those panels will grow from hundreds to thousands to hundreds of thousands of SNPs, and we'll be looking at them for hundreds of patients and thousands of patients and doing those experiments thousands of times.
BioInform: What do you mean when you say you will be scoring SNPs?
Pfost: I'll walk you through the lifecycle of a SNP. Not a lot of SNPs have gone through this life cycle, but this is going to be characteristic of the scientific and commercial activity for the next 10 years.
You get SNPs from primary sequencing. Then the first thing you want is to confirm that they're SNPs and actually run up the statistics. So, instead of looking at just a half dozen patients you want to look at 100 patients. What is this, is it really a SNP? Is it a 10 percent, 30 percent instance? Is it 1 percent or is it a mistake? Does it correlate to anything of any relevance at all? Or is there ethnicity attached? You do that to maybe 100 patients. You shouldn't have to do that using sequencers. You want to do that with a direct approach such as our SNP-scoring technique. That's what we call confirmation.
The next stage is association studies. With those you want to increase the statistics even further. And you may want to move from 100 patients to a few thousand patients and start to say, now this patient responded badly to a particular drug, or this patent had a predisposition to a particular disease, or this patient lived until she was 99 years old. What within these patients' genetic makeup can we look at and find where the correlations lie between individual polymorphisms and those medically important attributes?
SNP scoring is the rest of the story, but there's more richness to the story. First of all you want to say, well, is this SNP of interest at all? It may be a SNP, but is it of interest? What you do is come along with another piece of DNA, and it matches perfectly, it binds there. And then you come along with a special enzyme. You then extend this primer by one base. This one base either is or is not complimentary to that SNP. And that one base is modified such that later on when we shine light on it we get light out of it. That shows up in a square area on one of our chips where you will either get light or not get light, meaning that there either is or is not a SNP there. Depending on what we put down there, we can go and test each of the spots. We're going to be doing a whole bunch of them on 96-well plates so that each spot represents several hundred SNP tests, and that's where we get our throughput.
BioInform: What particular correlations will you be looking for, to start with?
Pfost: One of the correlations we're particularly interested in is people who have experienced adverse responses to taking a drug that has a toxicological effect or on whom maybe the drug didn't work. Why did that take place? Was it something to do with the way they metabolized the drug? Did it build up in their body and cause a toxic effect? Did they metabolize it too quickly? Did it bind someplace it wasn't supposed to bind or did it not bind at all? That variability is a big medical problem; it's an unmet medical need. Adverse drug responses are the fourth or fifth leading cause of hospitalization. It's underreported because you go to the doctor with congestive heart failure and if you end up dying in the hospital from a side-effect from a drug we don't know. We know you died of congestive heart failure.
What we're doing is finding those associations and patenting the use of drugs in situations. What that means is that for drugs already in existence that have known toxicological side-effects or other forms of idiosyncratic response we will be looking to uncover the genetic correlations and patenting the use of drugs where that genetic test has been applied.
BioInform: Are you more interested in patenting uses of drugs that already exist than in designing new drugs?
Pfost: Yes, primarily. If a big pharmaceutical company is in the developmental stage with a drug, and they're going to be doing clinical trials, this sort of work is very much of interest to them. It's just rather difficult to get their samples. You can't get an investigational drug very easily. You'd have to do your own clinical trials and that would be expensive. So, in that case, we'll be looking for pharmaceutical companies who want us to run their samples for them. We will be getting into collaborations where we would be benefiting from these associations and perhaps getting royalties from pharmaceutical companies or in some cases doing it on a service basis.
BioInform: This seems to respond to the concern that pharmacogenomics will shrink the market for blockbuster drugs.
Pfost: Yes. We can put a stop to that right now. I don't think that many people are thinking that way--it just sort of filters into the popular press--and I'll tell you why. If you have a drug that is good for a few hundred thousand people, sometimes the reason why it's a medium drug is because 10 percent of the people are having an adverse side-effect and the doctor doesn't want to prescribe it. If you could remove that 10 percent of the population so that 100 percent of the people who take it respond well, then that drug could become a blockbuster.
Although technically it's true to say this technology could enable what people call designer drugs or boutique drugs, it's not going to go in that direction. The reason is that everybody wants a blockbuster drug. They don't want to spend $400 million developing a drug that's only going to be for 10,000 people. This is all about keeping drugs on blockbuster paths, not paths to being a boutique drug.