Skip to main content
Premium Trial:

Request an Annual Quote

How Millennium Makes the Most of Microarrays


Millennium’s CTO Michael Pavia says the company employs its homemade DNA chips to streamline just about every area of the drug discovery process.

By Peter Gwynne


Michael Pavia has a mission. He wants to cut the cost of bringing new drugs to market by 50 percent. As chief technology officer of Millennium Pharmaceuticals, he is well placed to do so.

Pavia’s mission involves several strands of new thinking and advanced technology. One key factor is the use of microarrays. “In my early days in the industry we all had favorite genes, or targets as we called them,” recalls Pavia. Individual scientists concentrated on ‘their’ genes, poking and probing them in the effort to learn everything about their structure, character, and action.

Satisfying as that might have been for individual investigators, the ‘one gene, one experiment’ approach had little chance of providing a complete picture of how several genes work in harness and influence each other. Not surprisingly, the method now has the ring of pre-history. “We have the reverse situation,” says Bob Tepper, a physician who is Millennium’s chief scientific officer. “We have all the genes we need. Now we have to select them.”

Microarrays make selection of useful genes feasible, and Millennium has become one of the world’s most prolific users. Indeed, when it comes to applying microarrays to gene expression studies, the company probably holds the gold medal — it runs between 6,000 and 8,000 microarrays per month.

And unlike most pharma companies, Millennium insists on using arrays of its own manufacture, despite the availability of commercial microarrays from Affymetrix and others (see sidebar, page 23).

To be sure, Millennium isn’t shy about taking on do-it-yourself projects: it builds its own or modifies vendors’ equipment in many technology areas. But its DIY microarray efforts also reflect the fact that Millennium entered the field at much the same time as Affy and its competitors. Because microarray technology at the time was pretty rudimentary, “we decided to build our own,” says Pavia. “We continually looked at others’ microarrays and said that if a superior one came along we would use it. Nothing has come along yet.”

While Millennium has worked at the forefront of microarray development since the technology emerged, it was the idea of saving costs in drug development that stimulated its expansion of microarray usage. The project started in response to a query from Millennium CEO Mark Levin. “We found that a drug costs about $400 million to bring to market today,” recalls Pavia. “But we also found that it would cost only a quarter of that if everything went right in the process. The rest represents investments that fail.”

Allowing for a certain number of failures, the company decided to aim at improving its productivity in drug development by 100 percent over the typical industry figure, thus bringing down the total cost to $200 million. Such an approach, says Pavia, saves more than money. It should reduce by two to two-and-a-half years the 10-12 year time span now taken to develop, obtain approval for, and manufacture a new drug.

Pavia identified two routes to achieve the goal. “We decided to focus first on the $100 million successful project and ask how we can make it more efficient in terms of manufacture,” he explains. “By applying tools familiar to the automobile and other manufacturing industries we should be able to get the cost down to $80 million. Next we asked where the failures occur, and how we can bring their cost down by 60 percent, to $120 million.”

tips from toyota

Two years ago, to help start the process, Pavia made an unusual hire. He brought on board Helen Han, a professor of process engineering at Harvard Business School who had worked at Toyota but had no knowledge of the pharmaceutical industry. “We asked her to look at pharmaceutical discovery,” says Pavia. “She quickly showed that it’s not very efficient.”

“The biggest problem I identified at Millennium was that the interfaces were killing us,” says Han. “We had lots of pockets of excellence in individual groups, but when the handoffs had to go from one group to another, the problems occurred.”

Han got Millennium’s scientists to think in untraditional ways, such as prioritizing among their experiments. “By choosing the most important you can cut time and save costs,” explains Pavia. “She also helped to overcome scientists’ tendency never to kill a project.”

But her most important ideas came from examining why drugs that have shown initial promise fail to reach the market, as part of the effort to reduce the costs of failure by 60 percent.

Looking through the entire process of drug development, from gene discovery to marketing, Pavia and Han identified two particular weak spots: chemical optimization and clinical trials. The failure rate of both, it turned out, could be reduced through the application of microarray technology.

Screening for failures

Chemical optimization involves the control of processes like the absorption and metabolism of drug candidates, as well as their toxicology and general biopharmaceutical properties. The object is to avoid taking a drug candidate to the stage of chemical optimization only to find that it’s toxic to some organ. To do so, says Pavia, “you might take 20 substances that are liver-toxic and 20 structurally similar ones that aren’t, and see if you can identify a gene that’s involved in liver toxicity. We’re using microarrays not only to find the best targets but also to screen out the toxic compounds.”

The clinical trial stage of drug development is even more significant. “Half the dollars we spend on drug development are there,” says Pavia. “Certainly some compounds fail at that stage because the discovery team chose the wrong target. But trials can also fail because the wrong patients take part — patients who for some reason or other will not benefit from the drugs under study.

“Patients are now picked based on physicians’ recommendations,” Pavia continues. “We need to understand which patients we have and why those who fail do so.”

Here again microarrays offer help. The idea is to provide a cost-effective, less labor-intensive way of collecting comprehensive genotypic data on individuals, so that organizers of clinical trials can select the patients most likely to benefit from particular drug candidates and least likely to suffer side effects.

Millennium Predictive Medicine, the wholly-owned subsidiary of Millennium, is expanding the process, by using microarrays to identify genes associated not only with specific diseases but also with particular stages of those ailments. “We put together disease progression models,” says John Unger, chief technology officer of the Predictive Medicine division. “We need to get medical input as to how a disease such as breast cancer progresses, so that we can get samples of cells in each category. When we run samples from disease progression models in our arrays, we look for the differences between the genes expressed in normal tissue and those from different forms of the disease.”

The Microarray Man

Short, balding, and sprightly, Mike Pavia is a somewhat unlikely champion of a technology that resembles molecular biology on speed. He has no academic background in a biological discipline. Rather, he obtained his PhD in synthetic organic chemistry from the University of Pennsylvania. He does, however, have 18 years of experience in the pharmaceutical industry.

Pavia first joined Parke-Davis (now part of Pfizer) in Ann Arbor, Mich., as a chemist concentrating on the central nervous system. There he became involved in the first drug marketed to treat Alzheimer’s disease. “After about eight years,” he recalls, “I got a little disenchanted with drug discovery.”

Instead of leaving the industry, he stayed with Parke-Davis to set up a chemical technologies group. Through that group he made significant contributions to the discipline of combinatorial chemistry. “At the time, a brilliant synthetic organic chemist could make 40 compounds a year,” Pavia explains. “Now the number is tens of thousands.”

Two years later, Pavia’s career shifted in another direction. He moved to Cambridge to help start up Genesys, a biotech company based in large part on combinatorial chemistry. Genesys was soon bought by Sphinx, another biotech based in North Carolina that specialized in high-throughput technology. Sphinx, in turn, was sold to Eli Lilly, which appreciated its combination of skills.

Pavia worked for the new ownership for three years, during which time he had a scientific epiphany. “I realized that if you work in a single discipline it doesn’t solve the whole process problem,” he recounts. “If you’re not thinking of the entire process from gene to patient, you won’t help the patient whatever you do.”

As those thoughts were fermenting, a venture capitalist he knew encouraged Pavia to meet Millennium CEO Levin, a chemical engineer who was reaching conclusions similar to Pavia’s. “Mark asked me to take charge of the whole operation,” says Pavia.

It was an audacious move. “I knew almost nothing about genomics and informatics,” Pavia concedes. But he soon saw his lack of expertise as a strength rather than a weakness. “It’s important that you’re not set in your ways,” he points out. “Experts in general tend to make incremental improvements.”


Microarrays have been anything but an incremental improvement. They do for the study of genes what combinatorial chemistry does for new compounds.

Working successfully with microarrays demands expertise far beyond that of the traditional chemist or molecular biologist. “We have a ‘wet group’ of 50-plus people doing experiments and a ‘dry group’ of more than 100 informatics scientists,” explains Pavia. “We also have a project management group that coordinates the wet people, the dry people, and the scientists working on further advances.”

So effectively are Millennium’s microarrays generating data that Pavia faces a constant problem of finding people to interpret it. “We’re still improving the technology, especially in quality control and informatics analysis,” he says. “Our ability to generate high-throughput data has surpassed our ability to meaningfully analyze it.” Adds Tepper: “One should not underestimate the importance of storage, retrieval, and examination of the data.”

Using custom-designed software and other tools, the informatics group helps to slice and dice that data in ways different from the traditional approach to medical research. In the past, researchers used the descriptions of diseases as the basis for categorizing them into molecular pathways and mechanisms. “Now,” says Tepper, “we can stratify diseases very differently. We can reassemble them from molecular pathways into different groupings.”

The basic approach isn’t new. A team led by Stu Schlossman at Boston’s Dana Farber Cancer Center, using monoclonal antibodies, was able to expand the number of categories of lymphomas from two to many by identifying the markers on the cells’ surfaces. However, the use of microarrays improves that process several fold. “In the post-genomics world we’re doing the same sort of thing a lot more efficiently,” says Tepper. “We are converting diseases that may have very different descriptions into groups that resemble each other.”

The fundamental objective of this type of categorization is to discover new means of diagnosing and treating diseases. “We try to identify unmet medical needs,” explains Unger of the Predictive Medicine division. “For example, there’s not a screening test now to identify ovarian cancer. Early detection is only happenstance. By the time the cancer is usually identified it’s too late.”

To advance its microarray technology further, Millennium recently announced a collaboration with Swedish company Biacore International AB. This aims to use Biacore’s technology to study the suitability of new proteins, identified though research in genomics and proteomics, as drug targets. “We want to go beyond expression data and get protein data to put into the computational system,” says Pavia. “It’s a constant mindset here that anything we’re doing today is not good enough. We need to take it to the next level.”


Millennium’s Way With Microarrays


Why does Millennium choose to make its own microarrays when it has a choice of several commercial varieties? Two reasons, says Craig Muir, the company’s vice president of process technology. First, he points out, “the field is only a few years old. We’ve been working on microarrays for that entire time. When companies like Affymetrix came forward as vendors, we had small groups doing arrays.” Muir’s team devised a form of microarray suitable for the scientific group studying yeast genes. “But once we got into it,” Muir continues, “we realized that our technique was more reproducible and slightly more sensitive than any other method out there. So we’ve built new generations of the technology.”

Millennium’s microarrays, created by laying down stretches of DNA amplified by PCR on a nylon substrate rather than the more conventional glass, have also proved more robust and of higher quality than others, says Muir. They have earned plaudits inside and outside the company. “We use thousands of arrays per month, in some cases with tens of thousands of genes,” says Muir. “And people we almost couldn’t pay to use the technology a few years ago started beating our doors down to obtain it.”


The Scan

Two J&J Doses

Johnson & Johnson says two doses of its SARS-CoV-2 vaccine provides increased protection against symptomatic COVID-19, CNN reports.

Pfizer-BioNTech Vaccine Response in Kids

The Pfizer-BioNTech SARS-CoV-2 vaccine in a lower-dose format appears to generate an immune response among children, according to the Washington Post.

Chicken Changes to Prevent Disease

The Guardian writes that researchers are looking at gene editing chickens to help prevent future pandemics.

PNAS Papers on Siberian Dog Ancestry, Insect Reproduction, Hippocampal Neurogenesis

In PNAS this week: ancestry and admixture among Siberian dogs, hormone role in fruit fly reproduction, and more.