Need Proteins? Just Do It In Canada

---

By Mark D. Uehling, Bio-IT World

Shorts, sandals and a polo shirt. In autumn.
Maybe that is what Aled Edwards, PhD always wears to work. Maybe his wardrobe was designed to challenge American preconceptions of Toronto winters that last 11 months of the year.

Or maybe no one cares what you wear when you drive the cost of finding a sinewy, convoluted protein structure down to $125,000 — perhaps a quarter of the typical cost at less-efficient centers.

Edwards has a cross-appointment in the medical biophysics and medical genetics and microbiology departments at the University of Toronto (U of T) (Toronto, ON). He trained as a biochemist at McGill University (Montreal, QC) and Stanford University (Stanford, CA). But Edwards introduces himself in a self-deprecating manner as the CEO of Structural Genomics Consortium (SGC). He describes the SGC as a charity, as if it were researching his favorite disease of the week.

In fact, with a staff of 60 in Canada and 90 in Britain and Sweden, Edwards is doing pharma’s bidding on an industrial scale. There’s a twist: the non-profit SGC deposits all its 3-D protein structures into the public Protein Databank, a repository similar to those for genes and single nucleotide polymorphisms (SNPs).

No Bacteria Need Apply
Unlike the Protein Structure Initiative (PSI) effort in the U.S., which is basic research funded by the U.S. National Institutes of Health (NIH), the SGC works almost entirely on human proteins. The proteins on the SGC hit list are the body’s biological “target” enzymes into which a drug can dock. Such structures cannot be patented and thus, in current terminology, are of pre-competitive scientific interest.

With a budget one-third as large as that of the PSI effort, the SGC has three principal academic groups doing X-ray crystallography (versus 10 supported by the PSI). So its output is all the more impressive. “We can do it first and fast,” Edwards says. “We can do it cheaper than anybody.”

SGC has funding from the Wellcome Trust (London, U.K.), GlaxoSmithKline (Middlesex, U.K.), Genome Canada (Ottawa, ON), and other rarely co-operating Canadian government and academic agencies. It is working its way through a list of 2,000 proteins that bind to small molecules and are thus relevant to drug discovery. In 2004, the first two SGC centers to come on line were his own, in Toronto, and Michael Sundstrom’s lab in Oxford, U.K. A third, in Sweden, was added this year.

The SGC goals are ambitious: 50 structures in the first year (79 were actually finished), 100 in the second year, and 200 in the third. “We’re going to blow through our year two milestones,” Edwards predicts. Three of the 125 proteins structured by the SGC to date are located in cell membranes, some of the trickiest of all. The PSI has finished almost 700 unique proteins after a three-year head start.

Productivity Is All
Despite his academic credentials, Edwards is obsessed with productivity. “The main focus is the delivery of structures,” he says tersely. “There is no magic elixir or special machine.” Edwards himself is vigorous, blunt, and irreverent — a Mel Gibson for the life sciences.

Edwards once proposed an effort similar to the SGC to the Canadian Medical Research Council. “I got the lowest score ever on a grant. People giggled when you started talking about doing this genome-wide.” Edwards seems to enjoy goading American and European colleagues to pick up the pace. “What we are beginning to show is this idea that human protein structures are going to cost a lot more than bacteria is not the case,” says Edwards.

When forced to pick two structures of broad interest, Edwards names human 11ß-hydroxysteroid dehydrogenase type 1. It plays a critical role in conversion of cortisol and cortisone; so it may be relevant in obesity, insulin resistance, dyslipidemia and arterial hypertension. Edwards also expects to shortly finish all of the structures of the family of cytosolic sulfotransferases, which he terms “a great springboard for chemical biology studies.”

With neither remorse nor delay, Edwards spurs on his colleagues. Some experimentation with his recipe is permitted, but delays and departures from best practices are ruthlessly scrutinized. IT projects are kept small, contained and manageable. He’s kidding, mostly, when he says that “every dollar spent on bioinformatics is a wasted dollar. One has to have a foil that challenges every IT project. Otherwise they metastasize.”

In a pay-to-play model, SGC board members get to select and prioritize which proteins get structured next. The fee to join the board is $15 million Canadian. A Swedish consortium recently signed on, which will add Stockholm, Sweden’s Karolinska Institutet as a major center in the effort. A few U.S. collaborators are also assisting.

Eight Steps
In a few quick sentences, Edwards offers his own wry summaries of the Japanese, American, and European attempts to find protein structures — a notoriously intricate art that begins with X-ray crystallography. Edwards says that finding a structure typically has eight major steps, each requiring a tough choice — a fork in the road. But each region of the world has found its own obstacles on that road.

The Japanese invested in a lot of laboratory equipment. “They have more NMR machines than anyone on the planet,” says Edwards. “It makes your jaw drop.” The Americans are funding 10 major groups and pitting them against each other: “Here’s your money — and fight,” says Edwards. The Europeans lost themselves in protracted discussions. Says Edwards: “They went to Barcelona and Milan and talked about what they can’t do because they don’t have the money.”

In the U.S., the director of the PSI effort at the NIH, John Norvell, PhD says that he’s just received preliminary productivity data for the 2005 fiscal year, and it looks like the cost per structure is in the range of the Canadian costs: $100,000 or less. Just a year ago, in 2004, Norvell says the cost of one structure was $180,000.

Norvell notes that in the next five-year phase of his project, more resources will be devoted to four especially productive campuses, with six additional sites tackling still-unsolved challenges in protein expression, crystallization, and protein complexes. The Canadians, ever mindful of NIH dollars that find their way north of the border, are quick to say that some of the published techniques south of the border helped them get off to a fast start.

Still, Edwards can be acerbic about his corporate supporters. Each pharma claims to have identified its own obscure protein in the body and divined a novel role for that protein. In fact, says Edwards, in one therapeutic area after another, the molecular target is the same across the industry. “P38 kinase,” he says, citing one example. “They’re all working on the same kinase.”

Lost Professors
It is his fellow academics, however, that probably receive Edwards’s most probing analysis. They have a limited understanding of the costs of their own protein-structure research, he says. When asked for typical project duration or cost metrics, they cite the quickest, easiest proteins and omit the hard cases. In his effort, Edwards says, the SGC is now publishing 13-15 protein structures per month, close to his goal of 200 proteins annually. In some cases, the cost is $100,000 per structure.

Most academics in any country, he believes, have trouble making distinctions between enlisting students and employees. Bad postdocs drift away on their own, to be replaced by another class, another crop of white-coated laboratory laborers. But poor employees can be difficult to remove and harmful to morale and productivity. As a result, Edwards hires carefully. Concentrating his team members in three discrete centres has ensured a certain focus, he believes, in contrast to geographically dispersed efforts elsewhere.

Edwards wonders if some academics have been all but ruined for meaningful work in drug discovery. Tenure may offer too much latitude to try too many idiosyncratic approaches, some of which have proven to be ill-advised. Creative tweaks to the SGC process are allowed, but only for so long. “If you can’t show me in six weeks that this [new technique] is better, you’re not doing it,” he advises. “Don’t deviate from the path too much because we know it works.”

He is dismissive of ivory-tower attempts to generate 3-D structures through algorithms or in silico prediction. Such tools, for now, fail to show all the clefts and dynamics of the binding site in the necessary detail. “The utility of these models to do drug discovery,” Edwards says, “is useless.” The SGC remains an avid user of software, however, and cites modeling experts MolSoft LLC (La Jolla, CA) as a partner.

In many ways, there is an urgency to Edwards’s vision for protein structure productivity that parallels Craig Venter’s push to work as fast as possible with the tools at hand. In both cases, industry provided a gentle nudge forward. As with Venter’s relentless focus on the cost per base of DNA, Edwards believes the cost per protein structure will become a benchmark number that will help identify and refine best practices in the field. “We are not successful because we’re the smartest,” says Edwards. “We’re successful because we’re the best organized.”