Cell states and atlas analysis
Compare cell populations and state changes using single-cell expression matrices, sample groups and cell annotations to investigate development, immune responses or disease-associated processes.
AI virtual cell · Research
AI Virtual Cell
Explore cell states, regulatory mechanisms and perturbation responses using single-cell and multi-omics data, connecting computational models with experimental validation.
From measurements to models
Real data. Testable cell models.
AI virtual cell research uses measurements of real cells to build computational representations of cellular states through machine learning. Within defined species, cell types and experimental conditions, models can help investigate expression patterns, link regulatory information and explore responses to genetic or drug perturbations.
Research begins with a testable question: which data are needed, how cell states are defined, what changes are predicted and which independent data can evaluate the results. Model outputs support the generation and prioritization of hypotheses for experimental testing.
Research directions
Three directions for a defined research question
Virtual perturbation and response
With matched perturbation and control data, evaluate predictions of expression responses to genetic intervention or drug treatment, compare candidate interventions and prioritize follow-up experiments.
Multi-omics mechanisms and targets
Integrate expression, chromatin accessibility and other suitable omics measurements to connect candidate regulators with functional pathways and develop testable mechanistic hypotheses and candidate targets.
Data · Model · Validation
From data assessment to validation
- 01
Research objectives and data assessment
Define the research system, prediction task, available data and validation conditions.
- 02
Data organization and cell representation
Assess quality, organize sample metadata and represent cell states while considering batches and experimental conditions.
- 03
Modeling and baseline comparison
Select suitable models, compare with simple baselines and design evaluations that prevent data leakage.
- 04
Perturbations and interpretation
Explore perturbation responses, candidate pathways and regulatory signals within the scope supported by evaluation.
- 05
Validation and review
Test hypotheses with independent data or experiments and document applicability and follow-up research suggestions.
Data preparation
- Single-cell or single-nucleus transcriptomics: expression matrices, cell and sample annotations, experimental groups and quality information.
- Perturbation data: records of intervention targets, doses, time points, controls and biological replicates.
- Suitable multi-omics data: single-cell transcriptomics, ATAC-Seq, ChIP-seq, CUT&Tag, spatial transcriptomics, proteomics, metabolomics and other relevant data.
- Independent data for evaluation: validation samples, held-out experimental conditions, or public datasets suited to the specific research question.
Results and discussion
- Data quality and cell-state analysis results.
- Evaluation design, baseline comparisons and applicable conditions.
- Candidate responses, regulatory signals and associated visualizations.
- Reproducible analysis records and recommendations for experimental validation.
Research collaborations are scoped to the specific question. Data requirements, modeling approaches and deliverables are agreed in the project plan. Predictive performance depends on data quality and experimental conditions; conclusions require independent data and experimental validation.
Start your project
A plan built around your research question
Discuss samples, key comparisons and analysis objectives with our project team.