RESEARCH USE
Where CellTypist fits
Prepare a supported single-cell expression matrix with consistent gene identifiers, choose a built-in model that matches the species, tissue, and expected cell scope, or train a custom logistic-regression model from a reliably labeled reference. Export predicted labels, scores, and any majority-voting result. Record the model file and version, normalization, gene mapping, parameters, and software version, then evaluate predictions against known markers, expert annotation, and independent samples.
Research tasks
- Predict cell types with a reference model
- Train a custom classifier for a defined tissue or study design
- Compare predicted labels with clusters, marker genes, and expert annotations
What to evaluate before use
- A built-in model's tissue, species, and label system will not suit every dataset. Novel states, doublets, low-quality cells, and batch shifts can be forced into existing classes.
- Predicted labels and scores are not experimental confirmation of cell identity. Preserve model provenance and review results with markers, study design, and domain knowledge.
Verification note
This entry summarizes the resource's role without assessing scientific accuracy or endorsing its outputs. Features and terms can change; consult the official source before adopting it for consequential work.
Last verified: 2026-09-18
Source: official documentation ↗