Vertex Pharmaceuticals finds that the best applications are the ones scientists and IT can put together fast, then fix later.
By Kevin Fogarty
April 7, 2002 | In a business environment in which most small genomic research companies find niche roles and partnerships with big pharmaceutical companies, Vertex Pharmaceuticals Inc. plans to compete directly with Big Pharma.
In a research environment in which most companies focus on gene targets associated with specific diseases, Vertex studies entire gene families, hoping to draw conclusions about different diseases from closely related protein structures.
And, in a technology environment in which it can take years to get a custom-built software application perfected, Vertex prefers to let scientists and developers jointly define what the application should do, assemble a prototype that satisfies most of that need, then fine-tune as necessary.
"Our approach is to ask how we can build in a couple of days something that will give the scientists 80 percent of what they want, let them play with that prototype system and learn from it, then decide what features they truly need," says Mark Murcko, chief technology officer and vice president for the Cambridge, Mass.-based pharmaceutical company.
Such a fast-track strategy has created Gene Family Central, a homegrown application that started out as an informal knowledge management system. But after on-the-job customizations and additions, it has blossomed into a primary resource for most of the company's researchers.
Trying to anticipate all the features that make Gene Family Central so useful would have been almost impossible, Murcko says. "Scientists might sit down and make up a spec for some Cadillac information system, but to build that system might take six months. We try to put a prototype, usually Web- and Java-based, in front of users that satisfies 80 percent of their needs, then find out how further to refine it to capture the other 20 percent."
That practical approach to technology has been a key to the success — or at least the high profile — of Vertex, which was founded in 1989 by maverick Harvard and Merck-trained biochemist Joshua Boger. Boger believed that computer-aided research could enable a startup to build marketable drugs fast enough to compete with giant pharmaceutical companies.
The company, whose early years were profiled in Barry Werth's bestseller The Billion-Dollar Molecule, has had
|Vertex IT at a Glance
|Vertex opts for simple, straightforward technology, but not at the expense of performance or flexibility.
mixed fortunes over the years, including the suspension of development in September of VX-745, a rheumatoid arthritis compound that was pulled after animal testing showed adverse effects on the central nervous system. But Vertex has also launched a successful AIDS drug, the HIV protease inhibitor Agenerase, and has 15 other drugs in clinical or preclinical development, targeting infectious diseases including Hepatitis C, sepsis, autoimmune and inflammatory diseases, cystic fibrosis, Huntington's disease, and cancer.
That drug pipeline, one of the broadest of any biotechnology company making small-molecule, orally-administered drugs, is the result of a parallel discovery process in which the company identifies multiple targets within gene families and may design several drugs to address each. Doing so, however, involves processing enormous amounts of data, both from its own research and from outside sources. This is a prime responsibility of the company's bioinformaticians and one of the main reasons for the fluid application development process.
This approach, which is designed to get tools and data into the hands of scientists as fast as possible, also keeps Vertex from investing time and money developing applications for research that may change dramatically over the course of a year or less.
"If it takes you two days to construct a prototype, you can go back once a year and re-do the prototype," Murcko says. "We'd rather provide the application for immediate, tactile, real-time feedback so the scientists can move their projects forward. This is all about saving time. It's all about being as quick and efficient and nimble as you possibly can. That's what really provides the value in a pharmaceutical or biotech environment."
Vertex is not entirely unique in that endeavor, but most companies build only prototypes that way, then rebuild them in C or other languages, with a more stable infrastructure, says Mike Swenson, life science analyst at International Data Corp. (IDC).
"Rapid prototyping can be a good approach in a fluid environment like this because the science is actually very immature," says Swenson. "There's certainly some validity in being concerned about spending six or nine months on particular algorithms and then to say when you're using it that the science on which it was built has been superceded."
A Family Affair
Vertex' most successful example of this approach has been Gene Family Central, which presents assay data, information gleaned from scientific journals and conferences, patent filings, internal company meetings and memos, data on molecular modeling and interactions, exploratory research, predictive data, even competitive analyses.
The application was originally designed to consolidate data from a few sources for a few bench scientists,
"There are 3,000 to 5,000 genes that look like reasonable targets, so we want to be able to pick the best targets to work on. We pull in all the data to help you decide whether the target is interesting."
—Paul Caron, Vertex bioinformatics chief.
but has become a primary tool for researchers at Vertex. Scientists use it, among other things, to ensure their work is up to date with the latest findings from inside and outside the company. It also helps them track the company's research priorities and eliminate time-consuming targets that aren't likely to succeed.
"There are 3,000 to 5,000 genes that look like reasonable targets, so we want to be able to pick the best targets to work on," says Paul Caron, Vertex bioinformatics chief. "We pull in all the data to help you decide whether the target is interesting."
"It's a wonderful communications tool," says Keith Wilson, head of structural biology and the man primarily responsible for the research Gene Family Central is designed to support. "It gives any scientist in the company the ability to see the results of decisions within 24 hours of when they're made. You may not get all the story behind the re-ranking of targets in a gene family, but the combination of seeing the new ranking lets [bench scientists] know what upper management thinks of certain targets and lets them focus their efforts on the highest interest targets."
An organization might be investigating 500 targets at once, any one of which could suddenly become less appealing depending on the latest research findings — thus the appeal of Gene Family Central. The application also includes access to modeling and visualization data that lets scientists examine the conformation of the molecules they are designing and how well they match the structures of their protein targets.
Wilson adds: "At our fingertips, we can pull up the whole family, then pull up the closest neighbors of the family and ask how close they are. The ability to pull up the shapes of all these pockets and compare them to the target so we can be specific in the design of the molecules is one of the real powers of the database."
The database itself has enhanced scientific productivity so much that Vertex was able to file 19 patent applications, covering more than 100 chemical scaffolds, during the second half of 2001.
A Prototype Proven
Gene Family Central was created in the same way that most other Vertex applications were born, Caron says. About two years ago, Caron and a couple of colleagues conceived a database that would gather published scientific data, commercial data, and internal findings on a scientist's computer.
The resulting tool, built in Java using databases from Oracle Corp. (as well as other data formats and data sources), saved so much time for the scientists involved that Caron and Vertex's IT team eventually expanded it to include almost every area of the company's research.
Many applications at Vertex are built in Java, which allows them to run on many platforms, but an app's specific tools and languages — and the servers, operating systems and databases on which it runs — are trade secrets, says Paul Dupuis, director of information technology and information systems.
The key to the success of such an application, Murcko says, is not so much how it's built, but rather the human intelligence that selects what goes into it. A team of four curators with experience in drug development research works exclusively on culling data to determine what should be included. Staff experts also pitch in occasionally to identify more specialized information that should be included.
"There might be a million data points that come through a high throughput assay, but a bench chemist doesn't
|Vertex Products in the Pipeline
|Vertex' adept use of bioinformatics and its focus on exploring gene families have helped keep its product pipeline full.
want to look at those," Murcko says. "What we want to do is make sure we can boil that information down in a way that makes sure the rest of the organization can focus on high-value bits of information."
In addition to drawing on in-house experimental results and published research data, curators also interview Vertex staffers who have attended conferences for accounts of unpublished results and even the opinions of speakers at the conferences.
"We make available all the data we can," Caron says. "For example, we can help educate the bench scientist on diabetes, what other targets may be related to diabetes, and the results on a particular chemical scaffolding that he might need to know."
Such a level of integration, especially with an easy-to-use interface, is a goal for many biotech and pharmaceutical companies. But not many have achieved it to any significant degree.
Many large pharmas invest major time and money into building tools for their scientists, but their data integration issues are as much organizational and political as they are technical. "There are pools of integration in a lot of pharmaceuticals, but there are no larger processes for information sharing across silos," Swenson says. "That approach shows a lot of promise."
"Our goal was never to build a huge, bloated IT infrastructure," Murcko says. "We needed to make sure the needs of the scientists were met and that we could quickly move information around the company. You gain tremendous benefit in knowing what other people are doing, with no barriers or impediments to information."
Vertex's main advantage is speed, so any tool that allows it to move even faster will help keep it ahead of larger competitors like Merck and Pfizer that are trying to increase their own agility, says Michael King, a bioscience analyst for Robertson Stephens & Co. in New York. "They [Vertex] may not have the experience of some of the big companies, but they certainly can be aggressive."
Vertex is methodically analyzing the human genome, one gene family at a time, investigating structures to see if they're implicated in different diseases. If the company's researchers can understand how one protein structure relates to diabetes, for example, they will know more about how a similar structure in the same family relates to heart disease.
Most of the genetic structures within one family are similar, so the hardest part is untangling the first few structures, says bioinformatician Paul Caron.
"Organizing targets into gene families makes tremendous amounts of sense," says King. "That's where you get your synergies, that's where you're going to get the maximum leverage from your chemistry."
The company's practice of aiming several drugs at the same target may make up for VX-745, the failed rheumatoid arthritis drug. Vertex already has two similar compounds, VX-702 and VX-850, which do not cross the blood-brain barrier. Since the VX-745 setback last fall, the company has also filled out its lineup with an unrelated drug, VX-799, a caspase inhibitor designed to treat sepsis.
Continuing development, and replacement of failed candidates, bode well for Vertex's long-term prospects, in part because of its gene family-based research model, but mainly because of the large number of drug candidates it has in development. That puts it on par with Big Pharma, King says.
But King injects a dose of realism, saying, "When a company advertises 75 to 85 percent effectiveness, you've got to take that with a huge grain of salt."
Kevin Fogarty is a writer based in Sudbury, Mass. He can be reached at firstname.lastname@example.org.