Systems biology is...

In 2000, two pathologists decided to try a different kind of systems biology. Ellen Berg and Eugene Butcher were feeling frustrated with the slow progress of in silico approaches, so they developed a series of in vitro cell-based assays they say can recreate some of the complexity of in vivo interactions. They describe their platform, known as BioMAP, as systems biology, even though it lacks a hallmark of the field: computer modeling.¹

"I really didn't like the early descriptions of systems biology from Hiroaki Kitano and Lee Hood, where their definition of systems biology was, 'We have to measure all of the components over time and then we're going to build a model of the system,'" says Berg, who (with Butcher) founded BioSeek in Burlingame, Calif. To her, the definition should be less exhaustive: "Systems biology has to take advantage of the interactions and complexity of biology." Butcher and Berg argue that cell systems can mark the beginning of a rational drug discovery process.²

So far, the stakeholders in the field have not agreed on a set definition of systems biology. "When you talk about systems biology, it means a lot of things to a lot of people." says Lionel Carnot of Bay City Investors, a venture capital firm that has been with BioSeek for several years. "Every company you go to has a different definition."

BioSeek uses cell-based systems biology to screen potential therapeutics. Test compounds are added to one of twenty disease models, and a readout of 8-10 proteins provides the BioMAP, which is then compared to reference profiles of known compounds.

So far, BioSeek has forged collaborations with companies including GlaxoSmithKline and Dynavax, and it recently announced a $10 million collaboration with Amylin Pharmaceuticals to identify novel peptide therapeutics. Since 2000, the company has received an additional $19 million from traditional venture capital sources.

BioSeek's BioMAP System is designed to replicate cell and pathway interactions in human disease. The company has a library of 25-30 primary human cell types, from which they construct "systems" that represent specific disease models such as cancer and autoimmune diseases. The asthma model, for instance, uses a combination of endothelial cells and mast cells; one of the inflammation models uses endothelial cells, lymphocytes, and monocytes. Once the system is selected, BioSeek adds a drug or agent to modulate the cells and then measures a set of 8-10 readouts, which tend to be biomarkers, including cytokines, chemokines, lipid mediators, and other molecules involved in cell-cell communication.

With the BioMAP inflammation models, Berg has been able to distinguish between multiple kinase inhibitor mechanisms using only a handful of readout parameters. ¹ "For cell-based assays, there's no platform that can detect as many kinases as we can."

Ellen Berg and Eugene Butcher of BioSeek

Other companies, including Novartis and Pfizer, have similar technologies in-house, but these are largely much more simplified single cell-based assays for high-throughput applications. In some cases, Berg says, companies have developed one or two more complex models for a single project, but nothing that rivals the diversity at BioSeek. Richard Nelson, director of biomolecular screens at Boehringer Ingelheim, says he has been impressed by BioSeek's technology, particularly in a landscape dominated by in silico approaches. "It's a fairly straightforward system out of which they are able to get very rich data." He says it's "quite remarkable" that they can distinguish between kinase inhibitor mechanisms.

Not everyone working with systems biology agrees. "They're basically a screening company that set up a few assays," says Keith Elliston, president and CEO of Genstruct. "Calling that systems biology is a stretch." Though Elliston says he is impressed by the way BioSeek can read out multiple targets at once, at the end of the day, he's not sure what those data mean.

Bruce Gomes, head of mathematical modeling in the systems biology group at Pfizer Research and Technology Center, says he doesn't believe computer modeling is the hallmark of systems work. "There's nothing magical about computational methods," he says. "They're only successful if they deliver something that's useful to the drug discovery company."