Cell-state mapping
Compare cell populations and cell states using single-cell expression data and sample annotations. Identify patterns associated with development, immune responses or disease to guide further investigation.
AI virtual cell · Research
AI Virtual Cell
Connect single-cell and multi-omics data with AI to study cell states, regulatory mechanisms and responses to genetic or drug perturbations.
From measurements to models
From cellular data
to testable AI models
AI virtual cells use machine learning to build computational representations of cells from experimental data. Designed around the species, cell types and conditions of each study, these models help researchers investigate gene expression, explore regulatory relationships and predict cellular responses to perturbations.
Each project starts with a focused biological question. We define the data requirements, cellular states and prediction task, then plan how to evaluate the results using independent data. Model predictions help prioritize hypotheses for experimental testing, bringing computational analysis and laboratory research together.
Research directions
Three ways to explore cell biology
Perturbation response prediction
Use matched perturbation and control data to evaluate predicted gene expression responses to genetic interventions or drug treatment. Compare candidate interventions and prioritize follow-up experiments.
Mechanisms and target discovery
Combine gene expression, chromatin accessibility and other relevant omics data to investigate regulatory factors and functional pathways, generating testable hypotheses and candidate targets.
Data · Model · Validation
From data assessment to validation
- 01
Study scoping and data assessment
Define the biological system, prediction task and available data, and agree on an approach to validation.
- 02
Data preparation and cell-state profiling
Assess data quality, organize sample metadata and characterize cell states, accounting for batch effects and experimental conditions.
- 03
Model development and evaluation
Select models for the prediction task and benchmark them against simple baselines, keeping training and evaluation data separate.
- 04
Response prediction and biological interpretation
Investigate perturbation responses, candidate pathways and regulatory relationships within the conditions supported by model evaluation.
- 05
Experimental validation and next steps
Evaluate hypotheses with independent data or experiments, and identify the most useful directions for further research.
Data requirements
- Single-cell or single-nucleus RNA-seq: expression matrices, cell and sample annotations, study groups and quality-control information.
- Perturbation experiments: intervention targets, doses, time points, controls and biological replicates.
- Complementary omics: single-cell RNA-seq, ATAC-seq, ChIP-seq, CUT&Tag, spatial transcriptomics, proteomics and metabolomics, selected to suit the study.
- Independent evaluation data: validation samples, held-out experimental conditions or relevant public datasets.
What we deliver
- Data quality assessment and cell-state analysis.
- Model evaluation plan, baseline comparisons and guidance on applicability.
- Predicted responses, candidate regulatory relationships and biological visualizations.
- Reproducible analysis records and recommendations for experimental validation.
Projects are tailored to your research question. We agree on data requirements, the modeling approach and deliverables at the planning stage, with independent evaluation and experimental validation built into the research strategy.
Start your project
The right expertise for your next study
Tell us what you want to discover. We will help you define the samples, methods and analysis your study needs.