Skip to content

examples.md

Latest commit

 

History

History
361 lines (244 loc) · 8.06 KB

examples.md

File metadata and controls

361 lines (244 loc) · 8.06 KB

Examples

Natural language commands for spatial transcriptomics analysis.

Standard Workflow

A typical analysis follows this flow:

Load → Preprocess → Analyze → Visualize

1. Load Your Data

"Load /path/to/spatial_data.h5ad"
"Load my Visium data from /path/to/visium_folder"

2. Preprocess

"Preprocess the data"
"Normalize with log transformation and find 2000 variable genes"
"Preprocess with SCTransform normalization"

3. Analyze

Choose your analysis type below.

4. Visualize

"Show the spatial plot"
"Visualize CD3D expression on tissue"
"Create a UMAP colored by clusters"

Analysis Types

Spatial Domains

Identify tissue regions and niches.

"Identify spatial domains"
"Find 7 spatial domains using SpaGCN"
"Cluster the tissue into regions with STAGATE"
"Use Leiden clustering with resolution 0.5"

Common choices include SpaGCN, STAGATE, GraphST, BANKSY, and Leiden. See Methods Reference for the full supported list and defaults.

Cell Type Annotation

Assign cell types to spots or cells.

"Annotate cell types using the reference dataset"
"Transfer labels from reference with Tangram"
"Use scANVI for label transfer"
"Annotate with marker genes using CellAssign"

Common choices include Tangram, scANVI, CellAssign, SingleR, and mLLMCelltype. See Methods Reference for the full supported list.

Requires: Reference dataset with cell type labels (for transfer methods)

Deconvolution

Estimate cell type proportions in each spot.

"Deconvolve the spatial data"
"Estimate cell type proportions with FlashDeconv"
"Use Cell2location for deconvolution"
"Run RCTD deconvolution"

Common choices include FlashDeconv, Cell2location, RCTD, DestVI, and CARD. See Methods Reference for the full supported list and defaults.

Requires: Reference single-cell dataset with cell type annotations

Spatial Statistics

Analyze spatial patterns and autocorrelation.

"Analyze spatial autocorrelation"
"Calculate Moran's I for marker genes"
"Find spatial hotspots with Getis-Ord"
"Compute neighborhood enrichment"
"Analyze co-occurrence of cell types"

Choose the exact spatial statistic based on your biological question. See Concepts for how to choose, and Methods Reference for the full analysis-type list.

Spatially Variable Genes

Find genes with spatial expression patterns.

"Find spatially variable genes"
"Identify spatial genes with SpatialDE"
"Use SPARK-X to find spatial patterns"

Common choices include FlashS, SPARK-X, and SpatialDE. See Methods Reference for canonical defaults and supported values.

Differential Expression

Compare gene expression between groups.

"Find marker genes for cluster 0"
"Compare gene expression between tumor and normal"
"Find differentially expressed genes in domain 3"

Condition Comparison

Compare experimental conditions with proper statistics.

"Compare treatment vs control across patients"
"Find genes differentially expressed between conditions"
"Analyze condition effects stratified by cell type"

Requires: Sample/patient identifiers for pseudobulk analysis

Cell Communication

Analyze ligand-receptor interactions.

"Analyze cell-cell communication"
"Find ligand-receptor interactions with LIANA"
"Identify spatial communication patterns"
"Which cell types are communicating?"

Common choices include FastCCC, LIANA, CellPhoneDB, and CellChat (cellchat_r). See Methods Reference for canonical names, defaults, and species-specific settings.

Requires: Cell type annotations

RNA Velocity

Understand cellular dynamics.

"Analyze RNA velocity"
"Run scVelo velocity analysis"
"Use VeloVI for velocity estimation"

Common choices include scVelo and VeloVI. See Methods Reference for exact modes and accepted values.

Requires: Spliced and unspliced count layers

Trajectory Analysis

Infer developmental trajectories.

"Infer cellular trajectories"
"Calculate pseudotime with Palantir"
"Use CellRank for fate mapping"
"Compute diffusion pseudotime"

Common choices include CellRank, Palantir, and DPT. See Methods Reference for exact method names and requirements.

Pathway Enrichment

Find enriched biological pathways.

"Perform pathway enrichment analysis"
"Find enriched GO terms"
"Analyze KEGG pathway enrichment"
"Run GSEA on marker genes"

Common choices include Spatial EnrichMap, ORA, GSEA, ssGSEA, and Enrichr. See Methods Reference for defaults and parameter details.

CNV Analysis

Detect copy number variations.

"Detect copy number variations"
"Analyze CNV using immune cells as reference"
"Find chromosomal alterations in tumor cells"

Common choices include inferCNVpy and Numbat. See Methods Reference for canonical defaults and requirements.

Requires: Normal cell types as reference

Multi-Sample Integration

Combine multiple datasets.

"Integrate these three samples"
"Remove batch effects with Harmony"
"Combine datasets using scVI"

Common choices include Harmony, BBKNN, Scanorama, and scVI. See Methods Reference for the supported integration methods.

Spatial Registration

Align tissue sections.

"Align these two tissue sections"
"Register spatial slices for 3D reconstruction"

Common choices include PASTE and STalign. See Methods Reference for the supported registration methods.

Visualization Examples

Basic Plots

"Show spatial expression of CD3D"
"Create UMAP plot"
"Plot violin of marker genes by cluster"
"Generate heatmap of top markers"

Deconvolution Results

"Show cell type proportions on tissue"
"Create pie charts of cell composition"
"Visualize dominant cell type per spot"

Communication Results

"Show ligand-receptor dotplot"
"Visualize communication network"
"Plot top interacting cell types"

Spatial Statistics

"Show neighborhood enrichment heatmap"
"Visualize spatial hotspots"
"Plot Moran's I results"

Complete Workflows

Basic Spatial Analysis (5 min)

1. "Load /path/to/visium_data.h5ad"
2. "Preprocess the data"
3. "Identify spatial domains"
4. "Find marker genes for each domain"
5. "Visualize the domains on tissue"

Deconvolution Workflow (10 min)

1. "Load spatial data from /path/to/spatial.h5ad"
2. "Load reference data from /path/to/reference.h5ad"
3. "Preprocess both datasets"
4. "Deconvolve using the reference"
5. "Show cell type proportions on tissue"

Cell Communication Workflow (10 min)

1. "Load the spatial data"
2. "Preprocess with clustering"
3. "Annotate cell types" (or use existing annotations)
4. "Analyze cell-cell communication"
5. "Show the communication network"

Trajectory Workflow (15 min)

1. "Load the data"
2. "Preprocess the data"
3. "Analyze RNA velocity"
4. "Infer trajectories with CellRank"
5. "Visualize velocity streams on tissue"

Tips

Be specific when needed

  • General: "Analyze the data" → ChatSpatial can choose a reasonable default workflow
  • Specific: "Use SpaGCN with 7 domains" → ChatSpatial follows your requested method and settings

Chain commands naturally

  • "Load the data, preprocess it, and identify spatial domains"

Reference previous results

  • "Find markers for the domains we just identified"
  • "Visualize the deconvolution results"

Ask for help

  • "What methods are available for deconvolution?"
  • "How should I preprocess my data?"

Next Steps