In 1996, at a time when most biologists were still unaware of the promise mass computing held for their field, Keio University environmental information professor Masaru Tomita traded in his microscope for a computer. His goal: E-Cell, a comprehensive simulation of a single, entire cell with its complex network of metabolic pathways and products.
Tomita, who holds PhDs in computer science, electrical engineering, and molecular biology, has spent the last five years convincing people that not only is it possible to produce a virtual cell that can elucidate invisible metabolic changes, predict reactions to environmental changes, and generate hypotheses for further experimentation, “it’s something we can do … in 10 or 20 years.”
Tomita’s team first took on the parasitic bacterium Mycoplasma genitalium, whose genomic sequence — the smallest of all known organisms at 580 kb — was published in 1995. Using data describing the functions of the bacterium’s approximately 480 genes, researchers built a model, gene by gene, of a cell that could sustain itself in silico. Of the virtual self-sustaining cell’s 127 genes, 105 code for proteins and 22 code for RNA; it metabolizes glucose and produces ATP for energy, synthesizes proteins, and generates a phospholipid cell membrane. Researchers can knock out a single gene with the click of a mouse, and then watch the effects.
In 1999, Tomita’s team released the project’s E-Cell software online for beta testing (www.e-cell.org). They have since enabled the software for Windows (version 2, released late 2001) and aim to continually increase the sensitivity and flexibility of the software. Version 3, which is due to be released this year, will allow researchers to vary simulation engines — which construct model metabolic pathways based on differential equations, for example — to suit the needs of the specific pathway they are trying to model. Currently, users must choose a single engine for all of the pathways in the cell.
Meanwhile, the team has proceeded with increasingly complicated work: modeling complete cells. The core element of the model is the quantitative data on which it is built; and the team’s choices reflect the experimental data that is generally accessible in the literature. The researchers have focused on specific biological processes — bacterial chemotaxis, circadian rhythms, photosynthesis, cell cycle, and cell division — and a growing body of cell types. Subsequent models under development by the team include E. coli; human erythrocytes, neural, pancreatic, and myocardial cells; and rice.
There are parallel projects worldwide. Researchers at the University of Connecticut have developed the Virtual Cell, or V-Cell. DBsolve, BioSpice, and Gepasi are making similar efforts. In 2001, researchers at the University of Connecticut, including Leslie Loew and John Schaff, licensed their V-Cell model to Princeton, NJ-based Physiome in a drug discovery and development partnership. But Tomita is not interested in commercializing his software anytime soon. E-Cell aims to serve a specialist audience. Its true value, he claimed, derives from the extent to which it is an experimentally viable model. “The more people use it, the more valuable it is,” he said.
E-Cell is part of a larger effort at Keio University’s new Institute for Advanced Biosciences (IAB), and the software is used largely by students in Keio University’s year-old bioinformatics degree program, part of which is offered at the IAB campus, located less than a hour by plane from Tokyo. Tomita is also director of IAB.
He puts his students straight to work modeling, for which they must first comb the literature in search of a body of experimental data, both qualitative and quantitative. Lack of quantitative data is one of the biggest factors holding back the still-nascent field. Yet once they have accumulated the raw material — kinetic parameters, pH readings, osmotic pressure data, and so on — setting up the model program takes anywhere from two days to two weeks.
After debugging, the student has a workable “toy model.” Then the task of really understanding how it works begins. Tomita counsels his students to find and meet with scientists who specialize in specific pathways and deepen their own understanding of its workings. “It takes two years to really understand the biological aspects of the pathway,” he said.
One aspect that distinguishes the E-Cell effort is its existence within the broader scope of research efforts at IAB. The metabolome team, for example, is developing high-throughput measurement techniques for analyzing metabolite concentration and movement within cells. The equipment scientists currently use is only sensitive enough to make these measurements for samples containing hundreds of millions of cells. But the IAB researchers are hurrying to develop techniques that will accurately measure elements of various metabolic pathways at the level of a single cell. Another team is engineering knockout E. coli and other organisms.
The three units have also teamed up with other universities and private labs in a five-year, $15 million government-sponsored project to develop methods whereby scientists can study, design and, ultimately, engineer microorganisms with industrial potential. At IAB, the engineering team creates the organisms that the metabolome team measures. This data is given to the modeling team, which creates a model in silico, and then experiments with adjustments in order to design a better organism for the first team to engineer.
By this process, the scientists hope, they can tailor computer-aided design to the task of optimizing microorganisms for various industrial purposes and conditions. Reflecting their basic ingredient’s in silico “e-” status, they have dubbed the project “E2 Coli.”
Through such efforts Tomita and his collaborators, assistants, and students aim to bring about a new, IT-driven way of thinking about biological research.
Tomita wrote last year in Trends in Biotechnology that “it is still an open question as to whether or not it is feasible to construct a computer model of a whole living cell that is sufficiently sophisticated” to predict cellular behavior. The answer to that question may very well come from Tomita’s own lab — all from a 127-gene cell.
— Sara Harris, Tokyo