When will a 'virtual cell' (Arc Evo 2, Stanford ProtiCelli) accurately predict a cell's response to a perturbation unseen in training?
Chance it happens
67%Belief the predicted event happens at all — every “yes / by-a-date” outcome added together. “Other” is tracked as its own third slice.
Outcomes
Resolution criteria
A published result on a recognized benchmark where a virtual-cell model predicts an out-of-training perturbation response above an agreed threshold. Resolution horizon: 2031-12-31. Source: peer-reviewed publication, benchmark.
Evidence & context
Trials, publications and reports our research found on this prognosis. Each links to the original source.
This is an arXiv benchmark paper evaluating virtual-cell-style models on unseen-cell and unseen-perturbation generalization scenarios. It directly bears on the forecast because it assesses whether current models can predict out-of-training perturbation responses on recognized benchmark tasks and reports their current performance limits.
This is Arc Institute's official wrap-up of the inaugural Virtual Cell Challenge, a benchmark focused on predicting cellular responses to perturbations, including unseen perturbations. It bears directly on the forecast because it documents benchmark design, evaluation metrics, and top-performing virtual-cell models, though it is a challenge summary rather than a peer-reviewed publication establishing the resolution threshold.
This Arc Institute page describes the Virtual Cell Challenge dataset, leaderboard, and multi-metric evaluation framework for comparing virtual-cell models. It is relevant because the prediction resolves via a recognized benchmark, and this page provides evidence about the benchmark landscape and how out-of-training perturbation performance may be judged, even though it does not itself report a resolving result.
This is an industry research report analyzing virtual-cell models and explicitly identifying forward perturbation prediction as a key near-term application. It is not a peer-reviewed benchmark result, but it is directly on-topic background evidence about the technical milestone and its expected trajectory.
This appears to be a benchmark-focused scientific publication evaluating foundation cell models on post-perturbation RNA-seq prediction tasks. It bears directly on the forecast because resolution depends on a recognized benchmark showing accurate prediction of out-of-training perturbation responses.
This is a 2026 newsletter/news roundup covering Stanford's ProtiCell and related AI-for-biology efforts. It is not itself a peer-reviewed benchmark result, but it is directly about one of the named virtual-cell projects, so it provides indirect evidence about progress toward the forecasted milestone.