A bacterial cell, simulated molecule by molecule, lives its full life cycle on a computer

Cell, March 2026. DOI: 10.1016/j.cell.2026.02.009

Four hundred and ninety-three genes. One circular chromosome. A membrane packed so tightly with molecular machinery that the researchers had to render some of it invisible just to see what was happening inside.

A team led by chemistry professor Zan Luthey-Schulten at the University of Illinois Urbana-Champaign has built what amounts to a digital twin of a living cell, simulating its entire life cycle from DNA replication through protein production, metabolism, and division. The work, published in Cell, represents the first time anyone has captured a full bacterial cell cycle in a three-dimensional, dynamic simulation at nanoscale resolution.

The smallest possible test subject

To make the problem tractable, the team chose JCVI-syn3A, or "Syn3A" for short. Developed at the J. Craig Venter Institute, this is a minimal bacterial cell stripped down to carry only the genes necessary for life: replication, growth, division, and the core metabolic functions that keep the whole system running. With fewer than 500 genes on a single circular strand of DNA, it is about as simple as a free-living cell can get.

Simple is relative. The simulation still had to account for every gene, every protein, every RNA molecule, and every chemical reaction occurring inside the cell. The team drew on years of prior work, systematically modeling essential metabolic pathways and subcellular networks through a series of publications stretching back to 2018. Experimental data from collaborators at Harvard Medical School, including laboratories led by Angad Mehta and Taekjip Ha, provided critical measurements for validating the model.

When the chromosome became a bottleneck

Not all biological processes are created equal when it comes to computational cost. Graduate student Andrew Maytin discovered that simulating DNA replication was dragging the entire project to a crawl, nearly doubling the time needed to model the full cell cycle. His solution was to dedicate a separate graphics processing unit solely to chromosome replication while another GPU handled everything else.

This division of computational labor allowed the team to simulate the complete 105-minute cell cycle in six days of computer time on the Delta advanced computing resource at Illinois. That is still a long time, but it is fast enough to run multiple simulations with slightly different starting conditions and compare the outcomes.

Postdoctoral fellow Zane Thornburg, who worked alongside Maytin on the simulation, described the challenge of modeling simultaneous events in three dimensions across an entire cell. The final hurdle involved understanding how the membrane and the DNA coordinate their movements during cell division, a problem that required both to be modeled as dynamic, interacting systems rather than static structures.

Two minutes off from reality

The simulation's accuracy is striking. Across repeated runs with varying initial conditions, the simulated cell cycle matched the real-world timing within an average of two minutes. The model was repeatedly tested against experimental measurements, with discrepancies feeding back into refinements of the computational approach.

This is not an atom-by-atom simulation. The model averages the dynamics of individual molecules rather than tracking every atom, a necessary compromise given current computational limits. But the molecular-level resolution is sufficient to capture the timing and coordination of cellular processes in ways that cruder models cannot.

The practical value lies in the ability to observe many processes simultaneously. As Luthey-Schulten noted, the simulation can reveal what is happening in nucleotide metabolism at the same moment it shows DNA replication and ribosome assembly. In effect, it delivers the results of hundreds of experiments at once.

What a digital cell cannot yet do

The simulation has clear limitations. Syn3A is not a natural organism but an engineered minimal cell. Its behavior may not generalize to the more complex bacteria found in nature, let alone to human cells. The model also does not yet capture every aspect of cellular behavior; environmental sensing, responses to stress, and interactions with other cells are all beyond its current scope.

Scaling this approach to larger genomes presents an enormous computational challenge. A typical E. coli cell has roughly ten times as many genes as Syn3A, and a human cell has roughly 20,000 protein-coding genes. The computational cost would increase dramatically.

Still, the work establishes a proof of concept. A living cell's full cycle can be captured in simulation with meaningful accuracy. The researchers see this as opening a window on the foundations of living systems, one that could eventually be widened to include more complex organisms and more detailed molecular interactions.