API reference

Most of what follows is exported from the top level, so truecell.find_markers and truecell.markers.find_markers are the same object. The grouping below is for reading; it is not a package layout you need to know.

Two pages are the exception, and on those the import path shown is the one to use. The generics live on truecell.generics — truecell.generics.features(obj), not truecell.features(obj), which raises AttributeError. Seven of them are re-exported at the top level as well (create_truecell_object, create_assay_object, create_centroids, create_segmentation, create_fov, get_tissue_coordinates, as_graph); the other 66 are not. The loaders on Loading data likewise stay on their own modules: truecell.io.read_10x, truecell.datasets.pbmc3k, truecell.compat.anndata.as_anndata.

The docstrings are the primary source. Many of them record a specific decision about matching R — which of Seurat's two code paths a function follows, where a default was chosen to agree with Seurat:: rather than with the wider Python ecosystem, and the handful of places the two genuinely differ. Those notes are the reason this reference exists rather than a signature dump.

The map from Seurat

Seurat	truecell	Page
CreateSeuratObject, Seurat, Assay5	create_truecell_object, Truecell, Assay5	Objects
NormalizeData, FindVariableFeatures, ScaleData, SCTransform	normalize_data, find_variable_features, scale_data, sctransform	Preprocessing
RunPCA, RunUMAP, RunTSNE, JackStraw	run_pca, run_umap, run_tsne, jack_straw	Dimensional reduction
FindNeighbors, FindClusters, FindMultiModalNeighbors	find_neighbors, find_clusters, find_multi_modal_neighbors	Graphs and clustering
FindMarkers, FindAllMarkers, AggregateExpression, AverageExpression	find_markers, find_all_markers, aggregate_expression, average_expression	Differential expression
IntegrateLayers, FindIntegrationAnchors, MapQuery	integrate_layers, find_integration_anchors, map_query	Integration and mapping
AddModuleScore, CellCycleScoring	add_module_score, cell_cycle_scoring	Signature scoring
HTODemux, MULTIseqDemux, RunMixscape	hto_demux, multiseq_demux, run_mixscape	Demultiplexing and screens
LoadXenium, BuildNicheAssay, FindSpatiallyVariableFeatures	load_xenium, build_niche_assay, find_spatially_variable_features	Spatial
SketchData, LeverageScore, BPCells matrices	sketch_data, leverage_score, LazyMatrix	Working at scale
DimPlot, FeaturePlot, VlnPlot, DoHeatmap	dim_plot, feature_plot, vln_plot, do_heatmap	Plotting
Cells, Features, Idents, FetchData, LayerData	cells, features, idents, fetch_data, layer_data	Generics
Read10X, SeuratData::	read_10x, truecell.datasets	Loading data

Reading the signatures

Type annotations are resolved statically, straight from the source, so annotation-only imports guarded by if TYPE_CHECKING: still render and still cross-link. Three of them matter in practice — matplotlib.figure.Figure on every plotting function, Neighbor on as_graph, and Truecell on from_anndata — and all three are deliberate: they keep matplotlib optional and break two import cycles. See Fidelity.