Skip to Content

Speeding Drug Discovery

It takes years and millions to get a new drug to market. New techniques might burrow through the mountain of genome data and break the bottleneck.
October 1, 2001

The Woodlands, about 40 kilometers north of downtown Houston, is one of those planned communities that endeavors to provide virtually anything its residents might desire in an idyllic suburban setting. Among the Woodlands’ five villages are bike paths and hiking trails, parks and golf courses, a shopping mall of Lone-Star-State proportions and an arts pavilion, not to mention a hospital, schools-and what may be the world’s single largest genetically-engineered-mouse facility.

While the mice are not part of the careful plans of the community developers, they are key to the future of a local biotech company called Lexicon Genetics. Lexicon was founded six years ago by Arthur Sands and Allan Bradley, who were both at the Baylor College of Medicine at the time. Bradley is now director of the Sanger Centre in Cambridge, England, the largest European contributor to the Human Genome Project, while Sands stayed behind in Texas to run Lexicon and to tackle what he and Bradley assumed-rightly-would become the single most pressing issue in biology: Once the human genome is sequenced, what happens next? How is that copious information-the sequences of those 30,000 to 40,000 genes-transformed into medical therapies that will improve the lot of humankind?

“The genome encodes all potential drug targets for the pharmaceutical industry for all time,” says Sands. “It’s all there, encoded in the genes, which make proteins, which are the targets for drug discovery. Now we have the sequences. The big questions are, what do these genes do, and how do you mine the most valuable genes out of the genome for drug discovery?”

Lexicon’s answer is mice: 300,000 of them. Specifically, “knock-out” mice, in which a single gene has been targeted and disabled. Biologists have used such mice for over a decade to illuminate the functions of genes by studying how the mice develop without them. Lexicon, however, has managed to industrialize and automate the knock-out production process, rolling out the high-tech mice the way Detroit does automobiles. What once took months, Lexicon can now do in hours. The company is presently churning out 1,500 such genetically compromised mice a week, which is why it’s building its $40 million Woodlands mouse facility, the size of a few football fields, to hold them all.

It’s this kind of massive, grand-scale effort that could represent the future of drug discovery in the postgenome world. In the same way that it took an automated, factory-like effort to sequence the human genome in the first place, biologists are now automating some of their favorite research tools-from mice to fruit flies to worms-to make sense of the mountain of new information. This biology at warp speed is, in effect, the mission of the science that has become known as “functional genomics.” And for pharmaceutical companies facing the challenge of turning genomic information into actual drugs, functional-genomics tools are some of the hottest commodities around. To get their hands on as many of these tools as possible, drug firms are partnering with a number of biotech companies that, like Lexicon, promise to help unravel the genome’s mysteries.

Clogged Pipeline

For the past half-dozen years, drug industry experts such as venture capitalist Jurgen Drews, the former head of research at Hoffmann-La Roche, have been arguing that pharmaceutical companies are running perilously short on new drugs. Drews calls it “the innovation gap”: by his calculation, each pharmaceutical company needs to bring at least one new drug to market every year, and preferably two or more, to survive and prosper. Instead, drug firms have been averaging considerably less-.4 to .8 new drugs per year per company.

The sequencing of the human genome should, in theory at least, solve one aspect of the innovation gap. After all, the entire pharmaceutical armamentarium-all the drugs against all the varieties of human illness-is aimed at a grand total of less than 500 biological targets, severely limiting the number of diseases that can be treated and the strategies used to do so. The information contained in the human genome, say Drews and others, is likely to increase the pool of potential drug targets 10- or 20-fold.

But that bounty will come at a considerable cost. The process of transforming some biologist’s vision of a therapeutic gold mine into a new medication, certified by the U.S. Food and Drug Administration as safe and effective, is expensive and time consuming and getting more so every year. The latest estimate, from an analysis released in June by the Boston Consulting Group, suggests that pharmaceutical companies will spend about 15 years and $880 million for each novel drug that makes it to market.

First, a biological mechanism-a malfunctioning gene, for example, or the errant protein product of such a gene-has to be identified as critical to a disease process, and then that potential drug target has to be “validated,” proven to be truly relevant in the laboratory, whether in cells, in a test tube or in an animal model of the disease. Then drug candidates have to be created-perhaps small synthetic molecules or entire proteins-and screened to determine which of them can enter the body and the bloodstream and the relevant tissues and penetrate the cell in question, reaching the target and altering its function in some way that impedes the disease. This molecule has to be optimized for maximum efficacy with a minimum of side effects: it has to be tested for toxicity and perhaps reengineered; tested in live animals for safety; and finally, tested in humans-in perhaps thousands of patients-first for safety and then for efficacy.

There are many tight spots along this pipeline-and biotech firms are employing a host of new tools to open them up. But the rate of attrition is still staggering. For every drug that makes it to market, 50 or 60 candidates will have failed. And that’s in the pre-genome world, where the great majority of drugs are aimed at variations on those 500 familiar targets and based on well-known biological themes. With thousands of new drug targets, thanks to the Human Genome Project and other genomics efforts, it’s a whole new ball game-one with extraordinary new promise, and an entirely new set of risks. The upside, according to a study by the investment bankers at Lehman Brothers and the management consultants at McKinsey, will be pharmaceutical advances so profound that they “are nearly impossible to imagine, let alone predict.” The downside, says the analysis, could include a fourfold increase in the rate of attrition in drug development-200 drug candidates falling by the wayside for every single drug that makes it to market-and an astronomical rise in research costs. The report’s stunning conclusion: in the short term, the flood of new drug targets could be fatal to pharmaceutical companies. Over the next five years, the report warns, “the industry could go bankrupt by trying to innovate.”

The problem is that the sequenced genomes provide too many potential targets, but not enough biological understanding to go with them, a situation often referred to as “drinking from the fire hydrant.” With so many targets, entirely too many would-be drugs could be rammed down the pipeline and make it to human trials, only to fail after enormous expense. Understanding the functions of genes in order to identify the most promising drug targets is proving to be one of the best ways to help unclog the drug development pipeline.

Enter a host of functional-genomics firms that share a simple strategy: learn as much biology about these potential targets as technologically possible, and do it as quickly as possible. Of the many technologies used to those ends, says geneticist David Altshuler of the Whitehead Institute for Biomedical Research in Cambridge, MA, the most promising are those, like Lexicon’s mice, that will allow researchers to directly manipulate the functions of all of an organism’s genes one by one-and thus pinpoint those that play the salient roles in the causation, progression or prevention of disease.

A Lexicon for Biology

Before Lexicon came along, researchers had developed two techniques for creating knock-out mice, but both had severe limitations, explains Oliver Smithies, a pioneer of gene-targeting technology at the University of North Carolina at Chapel Hill. On one hand, researchers could target and disable a specific gene whose sequence was already known, but that process, as Smithies says, is “quite laborious, because you can only do one gene at a time.” On the other hand, genes could be mutated at random, which can be done quickly and with relative ease; but then the researchers wouldn’t know which gene had been mutated until they grew the animals to maturity-and perhaps not even then, if the absence of the gene was particularly subtle in its effect.

Lexicon’s advance is a technique that mutates genes at random but does so by using a virus to insert a known sequence of DNA into the genes. That sequence not only disables the mutated gene, creating the knock-out, but remains behind as a signpost to identify precisely which gene has been put out of action. With this technology, says Sands, Lexicon has managed to knock out 40 percent of the genes in the mouse genome, considerably more than all the other mouse researchers in the world have achieved in the last decade. Pharmaceutical and biotech companies, and even academic researchers, can sign on with Lexicon to access this extensive mouse library, which Millennium Pharmaceuticals, Bristol-Myers Squibb, Johnson and Johnson and a half-dozen others have already done.

Because the knock-out mice are meaningless without diagnostic technology to pinpoint the effects of the absent genes, Lexicon’s mouse facility in the Woodlands will house, not just 300,000 genetically compromised mice, but what Sands calls a “Mayo Clinic for mice” as well. This new center-where, of course, mice will be experimental subjects, not patients-will include a comprehensive radiology department, complete with MRI machines and CAT scanners designed to image the bones, organs and tissues of mice. It will have an immunology group to dissect rodent immune systems, and a neuroscience group, complete with a battery of behavioral tests, to study how the missing genes might affect brain development and behavior. It will have a developmental-biology group that will study how the absence of genes affects the development of the mice in utero, and a cardiology group, to look at how the absence of genes affects cholesterol, blood pressure, and heart and artery function. “Every medical department at Lexicon,” says Sands, “will be geared to study the function of genes in live animals and find those that are the most valuable for drug discovery.”

Biology on the Fly

What Lexicon is attempting to achieve with mice, a South San Francisco company called Exelixis is trying to do with fruit flies, worms and fish. These organisms have also established themselves over the years as workhorses in genetics labs, but Exelixis’s innovation was to put them to work on an industrial scale in order to elucidate the functions of genes and identify promising drug targets. As the Whitehead’s Altshuler puts it, “Flies and worms may not get diabetes, for instance, but they do sugar metabolism, and they do it pretty damn similarly to the way we do it. So you can find all the genes that affect sugar metabolism in the fly, find out if they’re relevant to humans, figure out their function and do drug discovery.”

Exelixis was founded by a trio of fruit-fly geneticists hoping to leverage the evolutionary conservation of genes and cellular circuitry that underlies these similarities-not to mention the arsenal of genetic tools honed by the decades of geneticists who have worked on flies, worms and fish. Since these organisms mature in just days or weeks, “You can rewrite their genetic code very quickly, so you can ask all the appropriate questions very quickly,” says Exelixis chief scientific officer Geoffrey Duyk. At Exelixis, those inquiries usually start with flies, and so Exelixis has amassed a collection of knock-outs covering most of the 13,000 or 14,000 genes of the fly genome. Using that library, for example, Exelixis researchers are investigating angiogenesis, the process by which new blood vessels are formed. This is one of the hottest areas of cancer research, because a tumor will spur the growth of blood vessels to feed its proliferating cells. Find a way turn off angiogenesis, the argument goes, and you can choke off the cancer. Fruit flies could help researchers tease apart the genetic underpinnings of angiogenesis, even though they don’t have blood vessels. What flies do have is a trachea: a system of branching vessels that carry air through the body. The development of the trachea is controlled by a process known as branching morphogenesis, which turns out to be the same process that creates blood vessels in humans.

The Exelixis researchers started studying branching morphogenesis in flies a year ago and expect to identify as many as 200 genes crucial to the process-all potential drug targets. The next step is to take these genes, find their counterparts in zebra fish, and then knock them out of the fish to see which are, indeed, involved in creating and maintaining blood vessels. Unlike flies, zebra fish do develop vasculature systems, just like mice and humans. And unlike mice and humans, zebra fish are, well, fish: their eggs develop outside the mother, and within 24 hours the body plan and all the organs are not only formed but visible, because the embryos and even the adult zebra fish are translucent. “You can follow in real time the development of the vasculature system in the embryo just by looking through a microscope,” says Felix Karim, who runs the Exelixis angiogenesis program.

The research to elucidate exactly what these genes do and how they cause disease then proceeds by a process Karim describes as “ping-ponging” between studying the genes of interest, or their absence, in flies, worms, zebra fish, mice and even humans-all done in parallel for maximum speed. The entire discovery process, from fruit fly to validation of a promising target, might take only a couple of months.
The same approach can also be used to identify the genetic targets of existing drugs or promising compounds, which is what Exelixis is now doing for Pharmacia, Bayer, Bristol-Myers Squibb and the National Cancer Institute. “Once we identify the molecular target,” says Duyk, “we can develop alternative compounds with the same target but which may have more optimal therapeutic and pharmacological properties.”

Cellular Insight

A still faster method for finding good drug targets, and even beginning to test potential drugs, forgoes model organisms and heads directly for the cells themselves. The technology is known as “cellular phenotyping,” where “phenotype” is the genetic lingo for how a particular gene manifests itself in an organism-from blue eyes, for instance, to a propensity for certain cancers. In a cell, those traits might translate into the presence or absence of a particular pigment or a tendency toward irregular size and shape. Because vast numbers of cells can be grown quickly and tested in parallel, cellular phenotyping not only speeds the process of elucidating a gene’s function, it opens the possibility of testing thousands or millions of potential drugs aimed at a particular target all at once. “You have to query the cell,” says Brent Stockwell, a chemist at the Whitehead Institute. “Ideally, you would start with diseased cells and then look for chemicals that would simply convert them back to normal cells.”

At Aurora Biosciences in San Diego, CA, for instance, researchers have created methods to measure in living cells the performance of the primary types of proteins that are known to be defective in diseases-and their accompanying cellular circuitry. Proteins called ion channels, for example, carry charged molecules back and forth across the membranes that surround cells and play key roles in maladies ranging from heart disease to diabetes to depression. That makes ion channels prime drug targets, but they are difficult to study because they only work when they’re actually embedded in the membranes of living cells.

Aurora’s technology to measure ion channel function uses a pair of fluorescent dyes that can be poured onto human cells growing in culture and then affix themselves to the cell membrane, one on the outside and one on the inside. Once attached, explains Paul Negulescu, senior vice president of discovery biology at Aurora, the dyes will respond to changes in the electric field across the cell membrane, which are controlled by ion channels. If the channels are conducting ions as they should be, then one of the dyes will glow. If they aren’t, then the other dye will turn on. To study a particular ion channel that might make a good drug target in treating a particular disease, says Negulescu, Aurora researchers can genetically engineer test cells to produce that channel, or in some cases work with cells that mimic the disease. They use an automated process to put the cells into thousands of separate wells on an “assay plate” and add the dyes; they can then test thousands of potential drugs per plate, perhaps 100,000 a day, looking for the ones that modulate the channel in a way that might cure the disease with a minimum of side effects.

Since Aurora first developed the technology, many of the major pharmaceutical companies have either licensed it or hired Aurora to develop custom variations of it. Aurora has also formed a consortium with Bristol-Myers Squibb, Merck, Pfizer and the Parke-Davis division of Warner Lambert to develop the technology further. In July, Vertex Pharmaceuticals acquired Aurora for about $600 million, in the hopes cell phenotyping will help bring more novel drugs into the pipeline and, says Vertex CEO Joshua Boger, help to more quickly unravel the biological workings of drug candidates already in development.

Ultimately, the more biological information pharmaceutical companies can acquire-and the faster they can acquire it-the more likely it is they’ll survive in the postgenomics world. “At the end of the day,” says biotech pioneer David Goeddel, who helped develop some of the industry’s first drugs, “those that have the best understanding of the biology are going to have the best success getting drugs out.”

Keep Reading

Most Popular

Large language models can do jaw-dropping things. But nobody knows exactly why.

And that's a problem. Figuring it out is one of the biggest scientific puzzles of our time and a crucial step towards controlling more powerful future models.

OpenAI teases an amazing new generative video model called Sora

The firm is sharing Sora with a small group of safety testers but the rest of us will have to wait to learn more.

Google’s Gemini is now in everything. Here’s how you can try it out.

Gmail, Docs, and more will now come with Gemini baked in. But Europeans will have to wait before they can download the app.

This baby with a head camera helped teach an AI how kids learn language

A neural network trained on the experiences of a single young child managed to learn one of the core components of language: how to match words to the objects they represent.

Stay connected

Illustration by Rose Wong

Get the latest updates from
MIT Technology Review

Discover special offers, top stories, upcoming events, and more.

Thank you for submitting your email!

Explore more newsletters

It looks like something went wrong.

We’re having trouble saving your preferences. Try refreshing this page and updating them one more time. If you continue to get this message, reach out to us at customer-service@technologyreview.com with a list of newsletters you’d like to receive.