Biotechnology · global
Helping Virtual Cells Remember Every Experiment: GenBio Unveils an Early Version of AIDO Cell
From genetic changes to subsequent drug treatment, AIDO Cell attempts to track how a series of interventions changes cells. The system currently covers only two human cell lines. Reproducing known drug mechanisms provides preliminary support, but whether new predictions can withstand experimental testing is the next hurdle.
Change a gene first, then add a drug, and a cell’s response may differ greatly from its response to the drug alone. For researchers, the challenge is not just predicting each step, but understanding how a previous treatment changes the starting point for the next experiment. On August 18, 2026, US-based GenBio AI announced its virtual cell system, AIDO Cell, which attempts to simulate these sequential experiments on a computer to help identify research directions worth committing laboratory resources to.
GenBio, co-founded by 2024 Nobel Prize in Chemistry laureate David Baker, describes the system as an early preview. According to the company, AIDO Cell uses a “world model” to maintain a cellular state that can change in response to interventions, bringing DNA, RNA, proteins, and overall cellular behavior into a shared simulation framework. Researchers can perform sequential operations and read predictions at different levels from the current state.
The research purpose of this design is to connect questions that were previously addressed separately: after a gene is suppressed, which proteins and pathways might change? What response would subsequent drug treatment produce? The system can also branch into different experiments from the same state to compare subsequent treatments. Chief Technology Officer Le Song explained to Lifespan.io that feedback across biological scales is intended to reduce the amplification of small errors as predictions pass from one level to the next; this remains a design goal that requires experimental testing.
The current preview supports K-562 and Hep-G2, human cell lines commonly used in biomedical research that originate from leukemia and a liver tumor, respectively. These models have accumulated extensive molecular data and experimental literature, making them suitable foundations for initial testing, but they cannot represent all normal tissues or patients’ cells. The company also plans to use new data through AIDO Foundry to adapt the simulator to other biological contexts.
For validation, early examples released by GenBio include reproducing imatinib’s known mechanism of action in leukemia cells, demonstrating how the system tracks drug effects across different biological levels. This provides a starting point for comparison with existing knowledge, but it does not yet answer the more critical question: can the model provide reliable answers in advance for experiments that have not yet been performed? The company says wet-lab validation of new predictions is underway.
“Whole-cell simulation” should therefore currently be understood as a research direction that integrates information across multiple levels, rather than a complete reconstruction of a human cell. Results involving cross-scale, sequential interventions still require independent evaluation and shared validation standards; performance in two tumor cell lines also cannot be directly extrapolated to efficacy or safety in humans. For drug development, the more concrete use at this stage is to narrow the range of hypotheses and candidate treatments to be tested.
Under the arrangements at launch, AIDO Cell was initially available to internal teams and a small number of early research collaborators, while a broader application program for academic, biotechnology, and pharmaceutical researchers was still being prepared. The tool’s value will ultimately depend on whether it can produce reproducible predictions in unfamiliar contexts and help researchers give their next experiment a stronger foundation.