Arc Institute opens registration for the Virtual Cell Challenge 2026, a competition for models that predict the effects of gene knockouts
Arc Institute opens registration for the Virtual Cell Challenge 2026, a competition for models that predict the effects of gene knockouts
On August 4, Arc Institute opened registration for the Virtual Cell Challenge 2026. Starting August 20, participants will predict how several cell lines respond when specified genes are knocked out. Arc will compare these predictions with withheld results from its own experiments. The top three teams will receive $100 000, $50 000, and $25 000.
A virtual cell is a program that takes a description of a cell's current state and calculates its response to an intervention. Under the competition rules, Arc will provide participants with control profiles for several cell lines, which are laboratory cell cultures that can be grown under stable conditions. Each profile shows gene activity before the experiment. Participants will then learn which gene must be knocked out.
The task tests whether predictions transfer across cellular contexts. Each model must predict how knocking out one gene will change the activity of other genes. Arc will compare the prediction with new experimental data that the model did not encounter during training. The same intervention can produce different responses in cells with different initial states. The competition will test whether each model captures these differences.
Last year's challenge was simpler. Participants worked with a single human embryonic stem cell line and approximately 300 genetic interventions. In its summary of the 2025 competition, Arc reported that almost every model performed worse than a simple baseline method when evaluated by mean absolute error. In 2026, Arc will extend the evaluation beyond a single cellular context.
The evaluation addresses a practical question: can a model identify hypotheses before an experiment that will agree with subsequent laboratory measurements? A prediction that agrees with those measurements could help a researcher decide which hypotheses to test next.
On August 20, Arc will release the dataset, task details, and evaluation rules. The final comparison dataset will be released on October 22, submissions are due by November 5, and the winners will be announced in late November.