YouTube Facebook LinkedIn Google+ Twitter Xinginstagram rss  

The PATH TO personalized medicine 
The tactics have changed, sometimes dramatically, but hints of the promise of pharmacogenomics are finally starting to trickle in from studies of asthma, cancer, and drug response.


In Part I of this series last month we examined some short-term strategies for exploiting genetic variations.

In Part II we now look at the platforms generating the first signs of real pharmacogenomics, and what those findings could mean to the industry as a whole.

Oct. 9, 2002 | The genome is equal parts Pandora's box and treasure chest: Lift the lid, and you're just as likely to reveal an ugly problem as a valuable find. It's been particularly rough on the frontlines of pharmacogenomics, where scientists are often heard lamenting, "It was harder than we expected." But promising new leads in asthma, cancer, and other diseases are bringing hope to the surviving, slightly nerve-wracked pioneers.

The groundwork of personalized medicine is forging ironclad connections between genetic variants and their corresponding diseases or drug responses. It is a massive undertaking. Researchers can't even agree on the total number of human genes, but this has not stopped scientists trying to discern which of the millions of variations in the genome hold the key to questions such as "Who gets heart disease?" or "Whom will this medication help?"

Finding the answers is essentially "an information flow problem," says Kari Stefansson, founder

Case Study: Genaissance's DecoGen Platform for Haplotyping 
Successfully applying SNPs to clinical studies requires satisfying both biologists' and statisticians' IT needs

Read More 
and CEO of deCODE Genetics in Reykjavik, Iceland. That's why many of the leading companies in the field are built around sophisticated bioinformatics platforms (see Case Study, right).

The technologies hitched to those platforms vary, but important themes in them are starting to emerge.

Automate and Integrate 
Millennium Pharmaceuticals Inc. of Cambridge, Mass., drew up the first blueprint for personalized medicine, applying cutting-edge informatics to knit together multiple, highly automated platforms. The company has had to tweak the model over the years, most notably by buying late-stage compounds and products because its own have not advanced quickly enough. But the mission and the informatics core are still there.

"We are focused on integrating genomics-based diagnostics and therapeutics with the ultimate vision of linking the right drug to the right patient," says John Maraganore, senior vice president of strategic product development at Millennium.

Millennium routinely uses genomic screening to separate out clinical trial participants whose

Gambling On Pharmacogenomics 
A Q&A with Klaus Lindpaintner, vice president of research and director of Roche Genetics and Roche Center for Medical Genomics

Read More 
unusual responses might cloud results. Many of its Big Pharma competitors, including GlaxoSmithKline, Pfizer Inc., and Roche, do this as well (See "Gambling on Pharmacogenomics," right).

But Millennium remains committed to its unabashed emphasis on genomics for generating new drug targets internally and its steadfast pursuit of pharmacogenomic diagnostics. Millennium and Roche, for example, are co-developing "diagnomic" tests for rheumatoid arthritis. And its alliance with Becton, Dickenson and Co. covers genomic markers for oncology. "These could be predictive either for drug therapy response or disease risk," Maraganore says.

The company's informatics needs have evolved, of course. But Maraganore notes: "Many of our original bioinformatics tools are still in place. For example, we still look for new targets in important structural families, such as kinases."

Millennium still carries out a lot of gene expression analysis but relies more on Affymetrix chips than the homemade arrays generated by its mammoth robot, Zeus. "The big bottleneck [is] not finding targets but understanding them," Maraganore says. Thus, pathways integration and knowledge management are the new priorities. Millennium has about 100 staffers working on software projects "running across the genome," Maraganore says.

"Knowledge management" is one of those fuzzy terms not typically associated with a hard-core science concern like Millennium. But the company learned the hard way that, though acquiring lots of data is good, the magic lies in its assimilation. "The key is building knowledge, not as a linear process but integrated across the value chain," Maraganore says. Millennium is attacking this problem at the software level with IT partner Ingenuity Systems Inc. And the company is also using its very corporate structure, establishing therapeutic area groups that integrate up and down the research and development pipeline, connecting divisions that might otherwise never formally interact.

Millennium largely inspired the high-throughput genomics generation, but its model has yet to produce a single pharmacogenomic product. Indeed, only one compound in the pipeline — MLN4760, which targets obesity — can even be called a genomics-derived drug. But copious cash and plenty of products in development count for something.

Relative Isolation 
Complex conditions like asthma, diabetes, heart disease, and obesity, caused by a mysterious mix of

Kari Stefansson, CEO and founder of deCODE Genetics, says their emphasis on informatics and statistics is one thing that sets deCODE apart, allowing them to generate many leads to genes linked to common complex diseases, which are the toughest to attack.
genetic and environmental factors, offer prime potential markets because they affect millions of people. But uncovering the genetic roots of such common conditions is demanding, tedious work, particularly when targeting a heterogeneous population such as the United States.

DeCODE has pioneered a fascinating approach to this problem. "Most other people are looking at the application of 'molecular tricks' to find new targets from the genome," Stefansson says. "We started with a core competence in genetics and by investing heavily in our own informatics system ... you get a foothold by understanding the nature of the disease."

The company has a unique resource for this task: data on 1,000 years of Icelandic genealogy. If an ancient mutation that increases susceptibility for a particular disease arose in a member of this small, isolated population, then genome linkage studies employing a battery of genetic markers, such as SNPs (single nucleotide polymorphisms) or microsatellites (repeating sections of DNA), should reveal the location of the errant gene.

"We realized from the start that we had to define the problem correctly," Stefansson says. "Genetic mutations are really just scrambled data, so we put an emphasis on bioinformatics and statistics." DeCODE is generating 15 million genotypes a month, Stefansson says, "but the key difference isn't the machines generating the genotypes, it's the software system that allows us to do the data mining."

The result is a growing list of novel genes linked to common disorders, including asthma, hypertension, and schizophrenia. For example, deCODE has uncovered a common asthma susceptibility gene and located two genes linked to respiratory allergies. Mapping the asthma gene required genome scans with 1,000 microsatellite markers. Nearly 1,000 people (about half of whom had asthma) from 175 families were scanned.

The next step is to create patient-tailored therapies and diagnostics, or new broadly acting drugs — depending on how common these genes turn out to be in other populations. The company has designed a gene expression test for predicting response to glucocorticoids, a common asthma treatment. Developing the test took more than 5 million data points, drawn from monitoring the expression pattern of 12,000 genes in patients with various responses to the drug. The test involves seven genes and is accurate nearly 90 percent of the time, according to deCODE. "We should have tests like this one on the market within two years," Stefansson says.

West Coast Offense 
Meanwhile, in San Diego, Sequenom Inc. is hinging its success to a set of more than 200,000 validated

Breaking Informatics Barriers 
Vast quantities of new data types need to be combined and analyzed now

Read More 
SNP assays developed for the company's popular MassARRAY genotyping platform. Sequenom started out as a tool provider, but was one of the first such companies to add a discovery arm. The 2001 merger with Gemini Genomics plc netted more than 20 million clinical data points from other isolated populations, including inhabitants of Newfoundland and the Labrador Islands.

Sequenom's prime resource when hunting for complex disease genes is a population of about 15,000 age-stratified, healthy Southern Californians. "We start by testing the hypothesis that the major risk factor in any complex disease is age," says Charles Cantor, Sequenom's chief scientific officer. SNPs that distinguish young people from the healthy elderly, he reasons, may point to age-related disease genes. The analysis is further simplified by focusing first on people of European descent.

"Genetics is a 100-year-old, systematic science that lets you associate genes to diseases," Cantor says. "Genomics is just a parts list." But access to the genome sequence accelerates studies, he says. "Thanks to the sequence, we can pick our SNPs intelligently. If we couldn't do that, we'd be running millions more genotypes for every study."

It has also influenced Sequenom's approach. "We like to do our genetics first, and then do genomics on a limited region. This is a more powerful technique," Cantor says. "With 100,000 to 125,000 assays, we believe we will find all the most interesting things."

Sequenom focuses on the most common SNPs, validating each finding in a different population to weed out statistical artifacts. Sequenom recently announced it has found more than 120 candidate disease genes, including some related to diabetes, osteoporosis, osteoarthritis, cardiovascular disease, anxiety, and depression, based on genome-wide screens using 25,000 SNPs. Its most advanced programs are in melanoma and breast cancer.

Genome Therapeutics Corp. also combines clinical genetics with high-throughput tools for

"Even if you have the chromosomal location, it is usually quite large. You need a platform that can look at lots of polymorphisms and pick the needle from the haystack."

Tim Keith, Genome Therapeutics Corp.

everything from positional cloning to genotyping. "We can find the location of the candidate genes, sequence the DNA, and collect the genotypes, all at high throughput if necessary," says Tim Keith, senior director of human genetics for the Waltham, Mass.-based company. "Even if you have the chromosomal location, it is usually quite large. You need a platform that can look at lots of polymorphisms and pick the needle from the haystack," she says.

As reported in Nature in early July, Genome Therapeutics discovered an asthma gene called ADAM33. The hardest part of that study was obtaining clean, relevant data. "High-quality clinical data is the foundation that drives the molecular genetics phase," Keith says. The project, carried out in collaboration with Schering-Plough Corp., involved multiple sites in the United States and the United Kingdom. Nurses were instructed to follow identical procedures and use identical instruments to ensure consistent practices.

"Finding this gene opens the door to elucidating a new pathway," Keith says. "This could lead to therapeutics that actually target the disease rather than the symptoms, as well as tests to pinpoint asthma sufferers even before they have symptoms."

Easy Targets 
Complex diseases offer a potent lure, but many companies would settle for any reasonable product — and hence the growing interest in targets considered within easier reach. That includes tests that could help personalize cancer therapy, since the drugs involved are some of the least efficacious and most likely to provoke side effects.

"Twenty percent efficacy can be a raging success for a chemotherapy," says R. Mark Adams, vice president of bioinformatics at Variagenics Inc. in Cambridge, Mass. Pharmacogenomic tools have provided an excellent window both into cancerous cells and the enzymatic pathways that influence drug response. "You can generate a very strong case for using these tools in cancer right now," says Adams.

TS, TP, and DPD Gene Expression vs. Survival - Variagenics points to evidence like this study comparing gene expression patterns in drug-responsive and unresponsive colorectal tumors to emphasize the potential value of pharmacogenomic tests. (TS=thymidylate synthase, TP=thymidine phosphorylase, DPD=dihydropyrimidine dehydrogenase.) 

The DNA in cancer cells undergoes myriad rearrangements, sometimes losing entire chromosomal sections. Variagenics' scientists use proprietary analytics to integrate that information along with gene expression patterns and genotypes, all captured at high speed. The result provides a much clearer picture of each patient's disease.

Variagenics is developing a suite of molecular diagnostics to guide colorectal cancer therapy. "We have a pipeline of prospective tests for each of the three major therapies — 5-fluorouracil, irinotecan, and oxaliplatin," says Jay Mohr, president and chief business officer. The company uses previously banked samples to find markers with diagnostic-caliber predictive power. "That usually works if we have big enough sample sizes and rigorous enough statistical modeling capabilities," Mohr says. The company hopes the power of those algorithms will give it a competitive edge. The next step toward an FDA-approved product is to develop a home brew, or analyte-specific reagent test that can be evaluated in a clinical trial.

In collaboration with Novartis Pharmaceuticals, Variagenics is looking for markers of response to two of Novartis' prospective prostate cancer therapies. One is Gleevec, a drug that has performed impressively against certain cancers. The other is PKI-166, an epidermal growth-factor receptor inhibitor still in development.

Riding the Third Wave 
The Japanese government is also interested in finding useful SNPs, particularly if they lead to lower health-care costs. It has chosen Madison, Wis.-based Third Wave Technologies Inc.'s Invader as the genotyping platform for what may be the largest SNP-related project in the world: producing more than 100 million genotypes per year while scanning 120,000 SNPs in about 1,000 patients.

"Phase one of the project is cracking the genetic basis of common diseases," Third Wave CEO Lance Fors says. "They've already found important markers for predisposition to asthma and arthritis and are in the midst of cardiovascular study. They will be tackling cancer next."

Drug-metabolizing genes are also part of the deal, and Third Wave has exclusive rights, outside Japan, to any genetic markers found in the study. Third Wave has already developed 10 reference lab-grade cardiovascular gene tests and plans to commercialize new tests as rapidly as possible. The company is also developing response tests for the colon cancer drug irinotecan in partnership with the drug's Japanese maker, Daiichi Pharmaceutical Corp. "A major goal of our collaboration with the Japanese government is making sure we can get those tests implemented so they are helping patients," Fors says.

Still other companies are trying to capture test markets for niche indications. For example, St. Louis-based DzGenes LLC is targeting kidney failure. "Everyone else is going after the biggest markets, like heart disease and oncology," CEO Terrence Kungel says. "We see a lot less competition in kidney disease, and we have an advantage in this arena." DzGenes has more than 10,000 DNA samples, more than half of which are from patients with renal disease; 336 SNPs appear to be linked to the progression of kidney disease, about 100 of which are tied to end-stage disease.

"Once someone gets on the slippery slope to kidney dialysis, they usually die within three to five years," Kungel says. "We need better tests to help us catch those patients before its too late."

Where will pharmacogenomics go from here? Will it "Balkanize" existing drug markets, reshape the drug prioritization process, revolutionize medicine, or merely spawn a wave of incrementally improved drugs, which are still effective in most people?

As those answers evolve, established companies must fight both scientific challenges and inclement market conditions, while new companies continue to stake their claims (see story below). "We are meeting our goals when it comes to revenue, but like everyone else in the market, we are in the toilet," deCODE's Stefansson says with characteristic bluntness. "But I don't think this tells us anything specific about ourselves as a company. Rather, I think it tells us something about the market."

Nevertheless, with patent expirations looming for many major drugs, pharmacogenomic companies have a broad window of opportun-ity. "The molecular diagnostic market is in the hundreds of millions [of dollars] now," Fors says. "It will be several billion within the next five years."

Millennium anticipates yet another kind of payoff in a shorter time span. "One or two hits are not that hard to generate," Maraganore says. "But building a pipeline that continues to deliver several new candidates year after year is something that's proved elusive to everyone. And that is what we're going to have by 2005, thanks to genomics." * 

DeCODE Genetics relies on Icelandic ancestry charts, such as the one shown above, to trace the roots of diseases such as asthma.

For reprints and/or copyright permission, please contact Angela Parsons, 781.972.5467.