Whole-cell simulation milestone
- Researchers simulated a minimal bacterial cell (JCVI-syn3.0) in 4D at molecular resolution, using GPU compute over several days. - The run took six days on A100 GPUs to simulate 105 real-time minutes of cellular activity. - Authors discussed code efficiency and prospects for scaling simulations to larger organisms, showing computational biology’s heavy compute demands (x.com).
A cell is a bag of molecules running chemical reactions in one tiny space, and researchers have now simulated that process for a minimal bacterium through an entire division cycle. (cell.com) The model follows JCVI-syn3A, a stripped-down bacterium from the J. Craig Venter Institute with about 500 genes on one circular chromosome. The team reported the work in *Cell* on March 9, 2026. (jcvi.org) Instead of treating the cell as a spreadsheet of reaction rates, the simulation places molecules in 3D positions and updates them over time, which is why the authors call it a 4D whole-cell model. It includes DNA replication, RNA and protein production, metabolism, growth, and cell division across a roughly 100-minute cycle. (cell.com) The choice of organism was deliberate. JCVI-syn3.0, unveiled in 2016, was built with 473 genes and a 531,000-base-pair genome as the smallest known self-replicating cell, and the newer Syn3A version restores a small set of genes to make growth and division more stable for experiments and modeling. (jcvi.org, sciencedirect.com) That matters because whole-cell modeling has chased a “virtual cell” for years, but most earlier models either covered only part of a cell’s chemistry or ignored where molecules sit inside the cell. A *Nature Methods* research highlight last week described this study as a 4D whole-cell model of a minimal cell, reflecting how unusual it is to combine full-cell timing with spatial detail. (nature.com) The compute bill was heavy even for a minimal organism. Illinois researchers said they used 15,000 graphics processing unit hours on the Delta supercomputer for 50 simulations, and each cell-cycle run took four to six days on two high-performance graphics processing units. (qcb.illinois.edu) The slowest parts were not always the ones biologists talk about most. Reporting on the team’s work, *Phys.org* said chromosome replication nearly doubled runtime in some versions of the simulation, which pushed the researchers to rewrite parts of the code for speed. (phys.org) The model is still not a literal copy of life. The Illinois team said it does not yet include every molecular interaction, even as it tracks genes, proteins, RNA molecules, membrane growth, and chemical reactions at nanoscale resolution. (chemistry.illinois.edu, qcb.illinois.edu) What the paper offers now is a test bed: a way to ask what happens if one gene is removed, one pathway slows, or one physical constraint changes before running a wet-lab experiment. For a field that has spent decades trying to turn cells into computable systems, the smallest self-replicating bacterium is where that effort is starting to look practical. (nature.com, nature.com)
