pwwang

cellcellcommunication

Infer ligand-receptor interactions and cell-cell communication networks from single-cell RNA-seq data using the LIANA+ framework. Identifies potential signaling events between cell types based on gene expression patterns and curated ligand-receptor interaction databases.

pwwang 21 4 Updated 4mo ago
GitHub

Install

npx skillscat add pwwang/immunopipe/cellcellcommunication

Install via the SkillsCat registry.

SKILL.md

CellCellCommunication Process Configuration

Purpose

Infer ligand-receptor interactions and cell-cell communication networks from single-cell RNA-seq data using the LIANA+ framework. Identifies potential signaling events between cell types based on gene expression patterns and curated ligand-receptor interaction databases.

When to Use

  • To identify ligand-receptor interactions between cell types
  • For systematic analysis of cell-cell communication networks
  • To understand cell crosstalk and signaling pathways
  • To compare communication patterns across biological conditions
  • To identify key signaling mediators in tissue microenvironments

Configuration Structure

Process Enablement

[CellCellCommunication]
cache = true

Input Specification

[CellCellCommunication.in]
sobjfile = ["path/to/seurat_object.rds"]  # Seurat (.rds, .h5seurat) or AnnData (.h5ad)

Environment Variables

[CellCellCommunication.envs]
# Method selection
method = "cellchat"  # Default inference method

# Cell type grouping
groupby = "ident"  # Column name for cell type labels (default: Seurat ident)

# Species selection
species = "human"  # "human" or "mouse"

# Filtering parameters
expr_prop = 0.1  # Minimum expression proportion (0.0-1.0)
min_cells = 5     # Minimum cells per cell type

# Statistical parameters
n_perms = 1000  # Permutations for permutation testing
seed = 1337      # Random seed for reproducibility

# Computational resources
ncores = 1  # Number of parallel cores

# Advanced options
subset = ""      # Expression to subset cells (e.g., "adata.obs.group == 'control'")
split_by = ""    # Column to run analysis separately and combine
assay = "RNA"   # Assay to use for Seurat objects

Available Inference Methods

LIANA+ provides multiple methods for cell-cell communication inference:

Method Description Magnitude Score Specificity Score
CellChat Mass-action-based communication probability lr_means cellchat_pvals
CellPhoneDB Permutation-based significance lr_means cellphone_pvals
Connectome Interaction-specific scoring - -
log2FC Log-fold change based - -
NATMI Network analysis - -
SingleCellSignalR Database-driven scoring - -
Rank_Aggregate Aggregates multiple methods - -
Geometric_Mean Geometric mean scoring - -

Default method: cellchat (recommended for most analyses)

LIANA+ Resources

Species-Specific Resources

# Human (default)
species = "human"  # Uses 'consensus' resource (CellPhoneDB, CellChat, ICELLNET, etc.)

# Mouse
species = "mouse"  # Uses 'mouseconsensus' resource

Available Resources (override with resource_name)

  • consensus (human default): Combines multiple curated resources
  • cellchatdb: CellChat database interactions
  • cellphonedb: CellPhoneDB interactions
  • mouseconsensus (mouse default): Mouse-specific consensus
  • icellnet, connectomedb2020, ramilowski2015, lrdb, and more

Configuration Examples

Minimal Configuration

[CellCellCommunication]
[CellCellCommunication.in]
sobjfile = ["path/to/seurat_object.rds"]

[CellCellCommunication.envs]
# Use defaults: method=cellchat, species=human, groupby=ident

Human PBMC Analysis

[CellCellCommunication]
[CellCellCommunication.in]
sobjfile = ["path/to/pbmc_seurat.rds"]

[CellCellCommunication.envs]
method = "cellchat"
species = "human"
groupby = "cell_type"  # Use annotated cell types
expr_prop = 0.1
min_cells = 10

Mouse Tissue Analysis

[CellCellCommunication]
[CellCellCommunication.in]
sobjfile = ["path/to/mouse_seurat.rds"]

[CellCellCommunication.envs]
species = "mouse"
method = "cellchat"
groupby = "seurat_clusters"
expr_prop = 0.15
min_cells = 8

Multi-Condition Comparison

[CellCellCommunication]
[CellCellCommunication.in]
sobjfile = ["path/to/combined_seurat.rds"]

[CellCellCommunication.envs]
split_by = "condition"  # Run separately per condition, then combine
method = "cellchat"
groupby = "cell_type"

Custom Cell Subset

[CellCellCommunication]
[CellCellCommunication.in]
sobjfile = ["path/to/seurat_object.rds"]

[CellCellCommunication.envs]
subset = "adata.obs.tissue == 'tumor'"
subset_using = "python"
method = "cellchat"

Common Patterns

Pattern 1: Full Interaction Network (CellChat)

[CellCellCommunication]
[CellCellCommunication.in]
sobjfile = ["intermediate/seuratclustering/SeuratClustering/sample.seurat.qs"]

[CellCellCommunication.envs]
method = "cellchat"
groupby = "ident"
expr_prop = 0.1
ncores = 4

Pattern 2: Disease vs Healthy Comparison

[CellCellCommunication]
[CellCellCommunication.in]
sobjfile = ["path/to/disease_vs_healthy.rds"]

[CellCellCommunication.envs]
split_by = "disease_status"
method = "cellchat"
groupby = "cell_type"
expr_prop = 0.1
min_cells = 10

Pattern 3: High-Stringency Analysis

[CellCellCommunication]
[CellCellCommunication.in]
sobjfile = ["path/to/seurat_object.rds"]

[CellCellCommunication.envs]
method = "cellchat"
expr_prop = 0.2  # Higher expression threshold
min_cells = 20    # More cells required

Dependencies

  • Upstream: SeuratClustering (required), CellTypeAnnotation (recommended for meaningful labels)
  • Downstream: CellCellCommunicationPlots (visualization: network, circos, heatmap, box plots)

Validation Rules

  • Species matching: Set species = "human" or species = "mouse" to match your organism
  • Cell type grouping: groupby column must exist in metadata; use CellTypeAnnotation or SeuratClustering results
  • Expression thresholds: expr_prop between 0.0-1.0; recommended 0.1 for human, 0.15 for mouse
  • Cell type resolution: min_cells minimum cells per type; recommended 5-10 cells per type

Troubleshooting

No Interactions Found

Solutions: Lower expr_prop (e.g., 0.1→0.05), reduce min_cells, check groupby column, verify species parameter

Species Mismatch Error

Solutions: Verify species matches organism, ensure gene symbols in correct format (human: uppercase, mouse: title case)

Slow Execution

Solutions: Increase ncores, reduce n_perms for permutation methods, use faster cellchat method

Memory Issues

Solutions: Reduce ncores, use subset to analyze specific cell types, merge rare cell types

Unexpected Cell Type Pairings

Solutions: Increase expr_prop, check cell type annotations, consider spatial context of data

Best Practices

  1. Run with CellChat first (default method provides good balance)
  2. Annotate cell types first using CellTypeAnnotation
  3. Validate expression thresholds based on data sparsity
  4. Compare multiple methods (cellchat, cellphonedb) when possible
  5. Interpret results in biological context (tissue structure, cell location)
  6. Always visualize with CellCellCommunicationPlots
  7. Document parameters for reproducibility

References