V3Cell

V3Cell is a vision-guided framework that constructs in silico surrogates of organoids directly from non-invasive brightfield microscopy, addressing a central problem in disease modeling and drug discovery: predicting how organoids respond to chemical perturbations. Existing "virtual cell" models largely operate at the single-cell level and produce static endpoint predictions derived from destructive assays. This leaves a gap at the organoid scale, where biological identity is defined by tissue-level architecture and continuous developmental dynamics rather than single-cell features.

To bridge that gap, V3Cell pairs a foreground-aware model that builds static virtual 3D cells with a temporal module that forecasts developmental fate and models fate-conditioned spatiotemporal trajectories. Critically, the framework requires no omics profiling and no fluorescent labeling, relying instead on ordinary brightfield image streams that are routine to acquire and non-destructive to the sample. This establishes a brightfield-based paradigm for organoid-scale perturbation prediction that can run alongside live culture rather than terminating it.

V3Cell was introduced in a June 2026 bioRxiv preprint from the Xinjiang Technical Institute of Physics and Chemistry, Chinese Academy of Sciences, together with the University of Chinese Academy of Sciences. The work is demonstrated across colon, stomach, and lung organoid lineages.

#Key Features

• Brightfield-only inputs: The framework operates directly on non-invasive brightfield microscopy, requiring no omics profiling and no fluorescent labeling, which preserves the living sample for continued culture.
• Foreground-aware virtual 3D construction: A foreground-aware model builds static virtual 3D cells that, per the authors, closely match real samples across distributional metrics, micro-texture, and lineage-specific morphometrics, with small reported effect sizes for most descriptors.
• Temporal fate prediction: A temporal module predicts developmental fate from as few as six early-frame observations and models fate-conditioned spatiotemporal trajectories that recapitulate real perturbation responses.
• Multi-lineage coverage: The approach is validated across colon, stomach, and lung organoid lineages rather than a single tissue type.
• Perturbation-response modeling: V3Cell targets organoid-scale responses to chemical perturbations, the regime most relevant to drug screening and disease modeling.

#Technical Details

V3Cell is composed of two coupled vision components. A foreground-aware model performs static virtual 3D cell construction by isolating organoid structure from brightfield frames, and a temporal module forecasts developmental fate and generates fate-conditioned spatiotemporal trajectories. The authors evaluate the static reconstructions against real samples using distributional metrics, micro-texture descriptors, and lineage-specific morphometrics, reporting small effect sizes for most descriptors as evidence of fidelity. The temporal module predicts developmental fate from as few as six early-frame observations and models fate-conditioned trajectories that recapitulate real perturbation responses. The framework operates entirely on brightfield microscopy, requiring no omics profiling or fluorescent labeling.

#Applications

V3Cell is aimed at researchers running organoid-based disease models and perturbation screens, where the ability to forecast developmental outcomes from early, non-destructive imaging can reduce reliance on terminal omics assays and fluorescent reporters. Potential use cases include early triage of compound effects in drug discovery, longitudinal monitoring of organoid development, and generating virtual surrogates for in silico experimentation across colon, stomach, and lung lineages. Because it consumes only brightfield streams, it can integrate into existing live-imaging workflows without specialized labeling.

#Impact

By moving virtual-cell modeling from the single-cell, endpoint regime to the organoid scale with continuous dynamics, V3Cell proposes a label-free, non-invasive route to perturbation prediction that aligns with how organoid identity is actually expressed at the tissue level. Its significance lies in coupling phenotypic 3D reconstruction with temporal fate forecasting from minimal early observations across colon, stomach, and lung lineages. The authors state that code and data are available at the project's GitHub repository. As a June 2026 preprint, its benchmark standing relative to established methods awaits peer review and independent evaluation.