The People's Biology

Erica P. Johnson Systems biologists envision a hulking database where all biological knowledge can be stored, freely accessed, and designed to interact. From it, researchers could easily extract data to construct virtual molecular pathway models working in their respective networks and in dynamic contexts of time, space, and various environmental cues. Hypotheses could be plucked like apples from the electronic tree of knowledge, and drug targets would fall like leaves. Some want to play out this tremendous vision, but they know it cannot be done at a single lab, by a single investigator. Members of Alliance for Cellular Signaling (AfCS), call for a new scientific world order--a shift toward socialist science. "It's going to be cool if we can get people to do it," says Alfred Gilman, the University of Texas Southwestern pharmacologist who has lead the government- and pharma-funded alliance since its 1999 inception. The group intends to collect mountains of new data on signaling in mammalian cells, and to generate a centralized database of what are called molecule pages. In conjunction with Nature Publishing, the AfCS intends to create nearly 3,000 author-curated, peer-reviewed web pages for single molecules. Currently, upwards of 800 so-called minimolecule pages link data from outside sources like GenBank and SwissProt, to expert-entered information, but the process is just beginning. The real difficulty, however, could lie in redirecting the scientific mind from individual to collective. The AfCS comprises more than 50 investigators from 20 academic and industrial institutions, and has garnered financial support from pharmaceutical companies such as Eli Lilly and Company and Johnson & Johnson. Gilman, who won the 1994 Nobel Prize for discovering G proteins asks, "Can we get the community to behave in a benevolent, interactive, cooperative manner? Only time will tell." DAS DATA To start, the scope of this $10 million-a-year program is small, but only relative to all biology. As the team was securing the first-ever "glue grant" (large-scale collaborative project awards) from the National Institute of General Medical Sciences (NIGMS), Gilman and others developed a plan to characterize signaling networks in two specific cell types chosen for uniformity and scientific relevance. Although no mammalian cells could meet all their criteria, they chose murine cardiac myocytes and B lymphocytes. Efforts are ongoing to produce standardized cell cultures. It has admittedly taken longer than hoped, but experimental data are beginning to trickle out for the B lymphocytes. Myriad Genetics is using yeast two-hybrid screens to expand the list of possible signaling partners. About 160 signaling proteins tested against a pool of 600 DNA-binding domains yielded more than 300 interactions. Other laboratories have been screening a set of 32 ligands, 18 of which have shown measurable changes in cyclic AMP or calcium concentration, phosphorylation of at least one of 10 signaling proteins, or expression pattern changes on a 10,000-gene microarray. The data appear on the Web as they are produced, but piecing together signaling networks will require serious computational finesse. Adam Arkin, a bioengineering professor at University of California, Berkeley, is helping develop the computer modeling programs to chew through B-lymphocyte data. While acknowledging the appeal of these cells for their relevance to human health, he still considers them "a systems person's nightmare." Having fine-tuned his methods on yeast screens, Arkin will now work on the AfCS's ligand screen. The next step will be to apply similar methods to the cardiac myocytes and to test the effects of different ligand combinations. According to Gilberto Sambrano, UC-San Francisco, and the AfCS's system committees coordinator, testing the effects of ligand combinations should reveal where signaling pathways potentially interact. To further focus efforts, researchers will initially concentrate on 100 to 200 networked molecules upstream of PIP3 accumulation. This focus was picked "in part to make it easier to get started and also have something that might be interpretable sooner," says Sambrano. Prompt interpretation is vital, says Hiroaki Kitano, who directs the Japanese government-funded ERATO Kitano Project. The AfCS, Kitano says, must quickly create exploratory models to get a better picture of what parameters will best shape the signaling networks. Some experimental parameters will require less attention, he says. Arkin, a Howard Hughes Medical Institute assistant investigator, agrees but says such investigations are not feasible with current data. "It's not yet of sufficient quality and quantity to drive theoretical predictions of what experiments to do next." Agreeing upon standardized cell lines and protocols must happen before experimental parameters can be assessed and tweaked, Arkin says. MOLECULAR MANIFESTO Shankar Subramaniam, bioinformatics director at UC-San Diego, says ambiguity clouds the experimental data: "There's too much empiricism and too much language." One ligand binding to a receptor may activate a protein slightly more, slightly less, change gene expression, or change calcium concentration by infinitesimal degrees. "Those things are pretty vague," says Subramaniam, whose efforts have mainly dealt with the long-awaited molecule pages. Yet, creating a universal language that is acceptable to an individual researcher is less a matter of dictating standards and more a matter of creating a valuable resource so that scientists will be drawn to it. "What we hope will happen is that people will start looking at the ontology and realize that it is providing useful knowledge and useful data and accept it, embrace it, and modify it as they see fit," he says. It has taken two years to create a molecule page framework that will comprise author entries, outside data sources, curation, peer review, and links to AfCS data. The first entries from molecule-page curators should appear in the near future, says Gilman. "We've just now sent out questionnaires to the members, asking them to please give us information on covalent modifications of signaling proteins.... This will form something of a test bed to see how many people will give us an hour. If they won't give us an hour, they won't give us the week that we'll need later." It is also possible that individual researchers will not buy into this all-for-one manifesto. "It's pretty much individual scientists, competition, who finishes first, who gets the grant, who gets the paper. And I think that's the way it is," says UC-San Diego signal transduction scientist Michael Karin. Unless fundamental changes are made to the funding structure, he predicts business as usual. "The essence of NIH-sponsored research is investigator-initiated research. That's what the R01 [grant] is all about." Marvin Cassman, UC-San Francisco, is the biochemist who headed the NIGMS when it awarded the alliance's glue grant. The weaknesses of investigator-initiated studies drove the decision to fund, he says. "It seemed clear that there were a number of problems that were substantially larger than could be dealt with at any single investigator's laboratory." Even if these problems remain unsolved, Arkin takes an optimistic view. "The good news is that the lofty goal is the 'z' in a very long alphabet. Along the way I think a lot of interesting data [will get] generated that will be useful to the single scientist." Yet, already two years in, Gilman says he's become self-conscious about the project. "I'm getting tired of talking about something that's still in the early stages. I just want to do it." Brendan A. Maher can be contacted at bmaher@the-scientist.com. Courtesy of BibleHistory.com ONE CELL, ONE VOICE Researchers creating computational models from their data continue to develop new software packages to fit their needs; because of this specificity, it is highly unlikely that any one application will emerge to perfectly fit every situation. Yet, researchers wishing to build more complete representations of living systems want to combine existing models. Programs that use specific, custom-designed languages often can't communicate, raising a problem of biblical ilk. Some have been developing standard model description languages to bring commonality to this Tower of Babel. In 1999, Hamid Bolouri, now at the Institute for Systems Biology in Seattle, and others began developing SBML (systems biology markup language), an XML-based language that would offer a common static representation of the dialects--a universal translator of sorts--describing biological models. Bolouri, with two computer scientists and a biochemist/programmer, worked under the leadership of the Japanese government-funded ERATO Kitano Project to provide a means of sharing models among multiple packages. One problem, says Bolouri, arises when researchers publish their models on the Web. "If the model is written in some home-grown language for home-grown computer software ... unless you are at that university [or] institute where this work was done, there's not much you can do with it." CellML, developed by researchers at Auckland University in New Zealand and Physiome Sciences in Princeton, NJ, similarly describes different models in a standard way to allow for easy import and export of information. Physiome's Fred Livingston, who worked on CellML, says both CellML and SBML will enable researchers to exchange models, put different parts together, and "hopefully be able to create some new insights." Bolouri's group has also been developing the Systems Biology Workbench, a so-called application integration environment into which a researcher can plug disparate tools for model analysis and have them run together. But as in the Tower of Babel story, pride, not language, is man's ultimate undoing. "Obviously, there are social issues as well as scientific issues," says Bolouri. "When you integrate your software with other people's software, there's always a fear that you lose some level of control." Bolouri says his group has been working to allay such fears by keeping software open-source, and both groups have been collaborating. Says Livingston, "Within a couple of years I think that the CellML groups and SBML groups will probably work with each other to bring the best aspects of each language together into one."

The People's Biology

Interested in reading more?

Become a Member of