Recent developments in artificial intelligence and 4D microscopy are paving the way for virtual cell models designed to accelerate drug discovery and reduce reliance on time-consuming laboratory experiments. According to Mirage News, researchers have utilized 4D lattice light-sheet microscopy—a technique that captures how structures like mitochondria move in three dimensions over time—to build advanced predictive models and digital twins.
AI Models and 4D Microscopy Advance Virtual Cell Simulation
Mitochondria form dynamic, interconnected networks throughout cells that rapidly split, fuse, and change shape based on cellular health. Because these structural shifts reflect cellular condition, they serve as markers of disease and can be used to test new treatments for conditions such as cancer, diabetes, Alzheimer’s, and pediatric mitochondrial disorders, as described in studies published in Mirage News.
Deep-Learning AI and Physics-Based Digital Twins
Researchers pursued two distinct approaches using 4D microscopy data. In one study, a deep-learning artificial intelligence model named MitoSpace was trained on 40,000 single-cell 4D movies of cancer cells treated with 25 different compounds. The model successfully predicted the energetic state of cells and distinguished between drugs, grouping them by mechanism with 75% accuracy compared to 56% accuracy when trained on flat 2D images.
For a century we have believed that mitochondrial form function; this shows the relationship is strong enough that a model can learn it without ever being shown the answer,
said corresponding author Johannes Schöneberg, a Roger Tsien Chancellor’s Faculty Fellow, associate professor in the Department of Pharmacology at UC San Diego School of Medicine, and in the Department of Biochemistry and Molecular Biophysics, according to Mirage News.
In a second study published in Mirage News, scientists constructed a physics-based digital twin of a living cancer cell by defining rules for how its organelles behave. When asked to simulate the effects of the drug nocodazole—which partially breaks down microtubules—the digital twin reproduced the reduced motion, fusion, and fission rates observed in real treated cells without altering any parameters.
AIDO Cell and the Path Toward Computable Biology
Expanding on the concept of virtual cellular simulation, California-based biotechnology company GenBio AI announced an artificial intelligence system called AIDO Cell in late August, according to The National. Researchers from Harvard, Stanhope, and Abu Dhabi’s Mohamed bin Zayed University of Artificial Intelligence (MBZUAI) contributed to the system.
The UAE embassy in Washington highlighted the development, noting that AIDO Cell simulates how a cell changes in response to experiments across various measurements using a “world model”. According to UA.NEWS, world models allow users to take actions midway through predictions and decode how real-world observations would unfold.
GenBio AI is actively working on a virtual cell bank that will allow researchers to download custom cells and run open-ended experiments, and the company is seeking partnerships to accelerate the process. Eric Xing, co-founder of GenBio AI and president of MBZUAI, emphasized the broader vision for the technology, stating via The National: To truly understand life, we must model and simulate it across every scale.