STAT-agent

Map single-cell reference annotations onto spatial transcriptomics data using Tangram deep learning alignment. Projects cell type labels from scRNA-seq reference to spatial spots/cells. Works for both cell-level and spot-level (deconvolution-like) data.

Analyze cell-cell communication using CellPhoneDB statistical method to identify significant ligand-receptor interactions between cell types via permutation testing.

Analyze cell-cell communication using LIANA+ to identify significant ligand-receptor interactions between cell types. (Recommended!)

Annotate cell types using unsupervised clustering, marker genes, and LLM-based annotation. Fast alternative to reference-based methods.

Annotate cell types in spatial transcriptomics data using scANVI transfer learning from a reference scRNA-seq dataset. (Recommended for cell data)

Perform cell type deconvolution (or annotation on spot) on spatial transcriptomics data (Visium spots) using RCTD with a single-cell reference dataset. (Recommended for spot data)

Reference-based Bayesian deconvolution of spot-level spatial transcriptomics using Cell2location. Two-stage model that first learns cell type expression signatures from scRNA-seq reference, then maps them to spatial spots. Provides uncertainty estimates. GPU recommended.

Ultra-fast reference-based cell type deconvolution for spot-level spatial data using FlashDeconv. O(N) complexity via random sketching — much faster than RCTD or Cell2location. Pure Python, no GPU needed.

Identify spatial cellular niches using Harmonics hierarchical model. Detects microenvironments based on cell type composition and assigns niche labels to cells.

Compute spatial statistics including Moran's I (spatial autocorrelation of genes), Ripley's K (spatial point pattern of cell types), co-occurrence analysis, and centrality scores. Uses squidpy for permutation-based spatial statistical testing on neighbor graphs.

Compute neighborhood enrichment z-scores to identify which cell types are spatially co-localized or depleted from each other's neighborhoods. Uses squidpy permutation testing on spatial neighbor graphs. Requires cell type annotations.

Identify spatially variable genes using SpatialDE Gaussian process regression. Decomposes gene expression variance into spatial and non-spatial components to find genes with significant spatial patterns. Uses official NaiveDE preprocessing workflow with Storey q-value FDR correction.

Infer cell developmental trajectories and pseudotime ordering using expression-based methods. Palantir uses diffusion maps and random walks for branching trajectories. DPT (Diffusion Pseudotime) is a lightweight alternative using scanpy. No RNA velocity data required.

Compare pathway / gene-set enrichment between two user-provided gene lists (typically markers of two cell populations, clusters, or conditions) and draw a mirrored bar plot — one group per side of a shared axis. Takes a dict of two gene lists in adata.uns['enrichment_genes_groups'] and tests each against a gene-set library via gseapy.enrich (Fisher's exact test + BH FDR). Supports MSigDB Hallmark, GO, Reactome, KEGG, and any other Enrichr library. Requires species specification.

Align two cell-level spatial transcriptomics slices using STalign. User provides matching landmark points between source and target slices. Computes initial affine from landmarks, then refines with LDDMM. Creates a new aligned slice with transformed x,y coordinates.

Correct batch effects across multiple slices using BBKNN (Batch Balanced K-Nearest Neighbors). Modifies the neighbor graph to connect cells across batches, enabling joint clustering and visualization. Lightweight and fast.

Correct batch effects across multiple slices using Scanorama panoramic stitching. Finds mutual nearest neighbors across datasets to learn a shared embedding, preserving biological variation while removing technical differences.

Align multiple spatial transcriptomics slices using PASTE (Probabilistic Alignment of ST Experiments). Uses optimal transport to find correspondences between spots across slices based on gene expression and spatial coordinates. Produces aligned coordinates for joint analysis.

Find differentially expressed marker genes between groups using scanpy rank_genes_groups with Wilcoxon test. The comparison target is flexible — between cell types, clusters, spatial domains, slices, ROIs, or any user-defined grouping.

Test whether a gene list is enriched for specific pathways or gene sets using Over-Representation Analysis (Fisher's exact test). Supports KEGG, Reactome, MSigDB Hallmark, WikiPathways, and GO databases via gseapy.

Integrate multiple spatial transcriptomics slices using Harmony batch correction. User specifies which slices to integrate and an optional reference scRNA-seq with cell type annotations. Harmony corrects batch effects in PCA space and stores corrected embeddings back into each slice. If slices lack cell type annotations and a reference is provided, KNN label transfer can annotate cells using the integrated embedding.

Find enriched Gene Ontology (GO) terms for a user-provided gene list. Takes a list of genes stored in adata.uns['go_genes'] and tests which GO biological processes, molecular functions, or cellular components are over-represented. Uses Fisher's exact test via gseapy with Benjamini-Hochberg FDR correction. Requires species specification (human or mouse).

Compute per-cell pathway activity scores using single-sample Gene Set Enrichment Analysis (ssGSEA). Scores each cell/spot for pathway activation, enabling spatial visualization of pathway activity across tissue. Supports MSigDB Hallmark, KEGG, GO, and custom gene sets.

Identify spatial domains in spot-level data using GraphST (Graph Self-supervised Transformer). Uses contrastive self-supervised learning on spatial transcriptomics graphs to learn domain-specific embeddings. Robust to noise and batch effects.

Identify spatial domains using STAGATE (Spatial-Transcriptomics Graph Attention Auto-Encoder). Uses graph attention networks on spatial neighbor graphs to learn spatially-aware cell embeddings, then clusters them into spatial domains. Works on both cell-level and spot-level data.

Identify spatial domains in spot-level spatial transcriptomics data using SpaGCN, integrating gene expression, spatial location, and H&E histology image features.