Skip to main content
Premium Trial:

Request an Annual Quote

Stanford Researchers Develop First Complete Computational Model of a Single Cell


Stanford University researchers have developed the first computational model of a single-cell organism that accounts for every molecular interaction that takes place in its life cycle.

The team, led by Markus Covert, an assistant professor of bioengineering and chemical and systems biology at Stanford, used data from more than 900 scientific papers and combined more than 1,900 experimentally observed parameters to model Mycoplasma genitalium — a sexually transmitted parasitic bacterium that has the smallest genome of any free-living organism with only 525 genes.

The method used to develop the model and a few of its real-world applications are described in a paper published in a recent issue of Cell.

The researchers expect the model to be a useful tool for hypothesis generation since researchers can use it to predict the phenotypic effects of genomic changes and then either prove or disprove the predicted effects experimentally.

In cases where experimental results disagree with the predictions, users can make changes to the model to fix these disagreements, Covert told BioInform. Such changes would further improve the model's predictive ability, and could reveal previously unknown biological insights as well as suggest new experiments for researchers to perform to validate in silico findings, he said.

Covert and his colleagues explain in the Cell paper that they decided to “divide the total functionality of the cell into modules, model each independently of the others, and integrate these submodels together."

The paper notes that this approach, which integrates multiple mathematical approaches, enabled them to address a factor that hampered previous efforts to construct whole-cell computational models — the fact that "no single computational method is sufficient to explain complex phenotypes in terms of molecular components and their interactions."

For example, efforts based on ordinary differential equations "were limited by the difficulty in obtaining the necessary model parameters,” while approaches that required fewer parameters, such as Boolean network modeling, had some underlying assumptions “that do not apply to all cellular processes and conditions,” the paper states.

As a result, “building a whole-cell model entirely based on either method is therefore impractical,” the authors wrote.

They ultimately grouped cellular processes into 28 submodules and then used a variety of mathematical methods, such as flux-balance analysis and Poisson distributions, to model a different cellular function per submodel.

The mini models were then combined to create a single model by first assuming that each one performs its particular function every second and then performing simulations in a loop in which the submodels run independently at each time step, “but depend on the values of [cellular] variables determined by the other submodels at the previous time step,” the authors wrote.

The paper also outlined how the model could be used to make predictions.

For example, the team used the model to predict DNA-binding protein interactions in M. genitalium. Among other results, the model predicted that 50 percent of the bacterium’s chromosome would be bound by at least one protein within the six minutes of the cell cycle and 90 percent within the first 20 minutes.

In another example, the simulations with the model indicated that 284 genes are essential for M. genitalium growth and division while 117 are nonessential — a prediction that agrees with previously observed gene essentiality with 79 percent accuracy.

The researchers also noticed that the length of individual stages in the cell cycle varied from cell to cell, while the length of the overall cycle was much more consistent. Using the model, they hypothesized that the overall cell cycle's lack of variation was the result of a built-in negative feedback mechanism.

Cells that took longer to begin DNA replication had time to amass a large pool of free nucleotides. The actual replication step, which uses these nucleotides to form new DNA strands, then passed relatively quickly. Cells that went through the initial step quicker, on the other hand, had no nucleotide surplus. Replication ended up slowing to the rate of nucleotide production.

Covert stressed that these findings remain hypotheses until they're confirmed by real-world experiments. However, "if you use a model to guide your experiments, you're going to discover things faster. We've shown that time and time again," he said.

'Two Notable Achievements'

Covert told BioInform that he first learned the methods used to develop the M. genitalium model while working as a doctoral student in Bernhard Palsson’s laboratory at the University of California, San Diego.

Palsson's lab focuses on the reconstruction of genome-scale biochemical reaction networks; mathematical analysis procedures for genome-scale models; and experimental verification of genome-scale models with a current emphasis on cellular metabolism and transcriptional regulation in E. coli, human pathogens, and organisms for environmental and bioprocessing activities.

His lab recently published a paper in Nature Communications describing the first in silico genome-scale model that combines metabolism and protein expression for Thermotoga maritime — a gram-negative bacterium that flourishes at high temperatures.

In a conversation with BioInform this week, Palsson said that the Covert et al. paper adds “two notable achievements” to existing research focused on building genome-scale models of metabolism, protein expression, and other cellular activities.

First, it builds a comprehensive model of a small genome that “puts together metabolism, protein expression, cell cycle and some regulation,” he said. Second, it is the first simulation algorithm that “computationally describe[s] all of these things simultaneously as a coherent whole,” he said.

He also noted that the approach described in the paper provides a useful tool for “explaining and understanding” cellular activity rather than a new approach for “prospective design” of genome-scale models, for which methods already exist.

Palsson also added that the computational challenges associated with integrating the different modules still need to be addressed.

For their next steps, Covert and his colleagues plan to experimentally validate some of the model's predictions. This will involve studying properties of some previously uncharacterized proteins in M. genitalium that affect activities like DNA replication, for instance, Covert told BioInform.

Covert also sees the tool as a “critical first step” towards synthesizing organisms that can produce new antibiotics or help clean up oil spills. “That’s something we are interested in pursuing more critically,” he said.

Filed under