Liana-Py-Complete Skill

Comprehensive assistance with LIANA+ cell-cell communication analysis framework for single-cell, spatially-resolved, and multi-modal omics data.

When to Use This Skill

This skill should be triggered when:

Data Analysis Tasks

• Analyzing single-cell RNA-seq data for ligand-receptor interactions
• Processing spatially-resolved transcriptomics data with spatial coordinates
• Working with multi-modal omics data (e.g., CITE-seq, metabolite data)
• Performing cell-cell communication inference across different conditions
• Running differential expression analysis for communication networks
• Building spatial neighborhood relationships between cells/spots

Specific LIANA+ Features

• Using individual CCC methods (CellPhoneDB, CellChat, NATMI, Connectome, etc.)
• Running rank aggregation to combine multiple methods
• Creating visualizations (dotplots, tileplots, circle plots)
• Processing ligand-receptor resources and custom databases
• Implementing MISTy models for spatially-informed multi-view learning
• Calculating bivariate metrics for spatial interactions
• Managing multi-sample data with tensor-cell2cell or MOFA+
• Working with metabolite-receptor interactions via MetalinksDB

Method Development

• Comparing different CCC inference methods
• Customizing ligand-receptor resources for specific organisms
• Implementing spatial analysis workflows
• Debugging LIANA+ pipelines and error handling
• Optimizing parameters for communication inference

Quick Reference

Core Setup and Data Loading

Example 1 (Basic imports and setup):

import liana as li
import scanpy as sc
import omnipath as op
import decoupler as dc

# Load example data
adata = sc.datasets.pbmc68k_reduced()

Example 2 (Multi-modal data setup):

import numpy as np
import pandas as pd
import scanpy as sc
import liana as li
import mudata as mu
from matplotlib import pyplot as plt

# Load CITE-seq data
prot = sc.read('citeseq_prot.h5ad')
rna = sc.read('citeseq_rna.h5ad')

Running Cell-Cell Communication Methods

Example 3 (Individual method - CellPhoneDB):

from liana.method import cellphonedb

cellphonedb(adata,
            groupby='bulk_labels',
            resource_name='consensus',
            expr_prop=0.1,
            verbose=True,
            key_added='cpdb_res')

Example 4 (Rank aggregation across methods):

li.mt.rank_aggregate(adata,
                     groupby='bulk_labels',
                     resource_name='consensus',
                     expr_prop=0.1,
                     verbose=True)

Resource Management

Example 5 (Working with ligand-receptor resources):

# Show available resources
li.resource.show_resources()

# Select specific resource
resource = li.resource.select_resource('consensus')

# Generate custom LR geneset
lr_progeny = li.resource.generate_lr_geneset(lr_pairs, progeny, lr_sep="^")

Visualization

Example 6 (Creating dotplots):

li.pl.dotplot(adata=adata,
              colour='lr_means',
              size='cellphone_pvals',
              inverse_size=True,
              source_labels=['CD34+', 'CD56+ NK'],
              target_labels=['CD34+', 'CD56+ NK'],
              filter_fun=lambda x: x['cellphone_pvals'] <= 0.05)

Example 7 (Customizing plots with plotnine):

import plotnine as p9

(my_plot +
 p9.theme_dark() +
 p9.theme(strip_text=p9.element_text(size=11),
          figure_size=(7, 4)))

Spatial and Multi-Modal Analysis

Example 8 (MISTy models for spatial data):

from liana.method import MistyData, genericMistyData, lrMistyData

# Create and run MISTy model
misty = MistyData(adata, intra_groupby='cell_type', extra_groupby='spatial')
misty.run_model()

Example 9 (Multi-view tensor analysis):

# Convert LIANA results to tensor for cell2cell analysis
tensor = li.multi.to_tensor_c2c(adata,
                                sample_key='sample_id',
                                score_key='magnitude_rank',
                                liana_res=adata.uns['liana_res'])

Utility Functions

Example 10 (Data transformation):

# Extract obsm to new AnnData
new_adata = li.utils.obsm_to_adata(adata, obsm_key='spatial')

# Handle negative values
positive_data = li.utils.neg_to_zero(data_matrix, cutoff=0)

# Interpolate data to common space
interpolated = li.utils.interpolate_adata(adata1, adata2)

Key Concepts

Core LIANA+ Components

• Ligand-Receptor Methods: Statistical approaches to infer CCC from transcriptomics data
• Resources: Prior knowledge databases of ligand-receptor interactions (consensus, CellPhoneDB, etc.)
• Magnitude Score: Strength of interaction (e.g., mean expression)
• Specificity Score: How specific an interaction is to particular cell type pairs
• Rank Aggregation: Combining multiple methods using Robust Rank Aggregation (RRA)

Data Structures

• AnnData: Standard single-cell data structure with .X, .obs, .var, .uns
• MuData: Multi-modal data structure for multiple omics layers
• Spatial Connectivities: Neighborhood relationships between cells/spots

Analysis Types

• Steady-state CCC: Standard ligand-receptor inference from single-cell data
• Spatially-informed Analysis: Methods incorporating spatial coordinates
• Multi-modal Integration: Analysis across different data modalities
• Differential Analysis: Comparing communication across conditions

Reference Files

This skill includes comprehensive documentation organized in references/:

api.md (45 pages)

Core API documentation for all LIANA+ functions

• Method signatures and parameters for all CCC inference methods
• Utility functions for data transformation and spatial analysis
• Complete parameter descriptions and return values
• Best for: Looking up specific function syntax, parameter details, and usage examples

getting_started.md (4 pages)

Introduction and decision tree for choosing analysis methods

• Framework overview and capabilities
• Decision tree for spatial vs non-spatial analysis
• Installation and basic setup instructions
• Best for: New users getting started with LIANA+ and choosing appropriate methods

tutorials.md (Comprehensive tutorials)

Step-by-step walkthroughs of common analyses

• Steady-state ligand-receptor inference with detailed examples
• Multi-modal data integration workflows
• Spatial analysis and MISTy model implementation
• Best for: Learning complete workflows and understanding analysis pipelines

other.md (2 pages)

Additional information and changelog

• Version history and new features
• Citation information and acknowledgments
• Best for: Checking version compatibility and proper citation

Working with This Skill

For Beginners

1. Start with getting_started.md to understand the framework and choose your analysis approach
2. Follow the basic tutorials for steady-state ligand-receptor inference
3. Use the Quick Reference examples to get started with common tasks
4. Focus on individual methods first before moving to advanced features

For Intermediate Users

1. Explore the API documentation for specific function parameters and advanced options
2. Try multi-modal analysis if you have CITE-seq or spatial data
3. Experiment with different visualization options to best present your results
4. Use rank aggregation to combine multiple methods for more robust results

For Advanced Users

1. Customize ligand-receptor resources for specific organisms or interactions
2. Implement spatial analysis pipelines with MISTy models and bivariate metrics
3. Build multi-sample workflows using tensor-cell2cell or MOFA+ integration
4. Optimize parameters for specific data types and research questions

Navigation Tips

• Search by function name in the API documentation for quick reference
• Check the tutorials for complete workflow examples
• Use the decision tree in getting_started.md to guide analysis choices
• Refer to method details in tutorials for mathematical formulations and assumptions

Resources

references/

Organized documentation extracted from official sources. These files contain:

• Detailed explanations of LIANA+ concepts and methods
• Complete code examples with proper syntax highlighting
• Mathematical formulations for statistical methods
• Links to original documentation and citations

scripts/

Add helper scripts here for common automation tasks such as:

• Batch processing of multiple samples
• Custom visualization functions
• Data preprocessing pipelines

assets/

Add templates, boilerplate, or example projects here:

• Example analysis notebooks
• Custom resource files
• Configuration files for different analysis types

Notes

• This skill was generated from official LIANA+ documentation (version 1.6.1)
• Reference files preserve the structure and examples from source documentation
• Code examples are extracted from real tutorials and API documentation
• Mathematical formulations and method details are preserved from original sources
• LIANA+ is part of the scverse ecosystem and integrates with AnnData/MuData objects

Updating

To refresh this skill with updated documentation:

1. Re-run the scraper with the same configuration
2. The skill will be rebuilt with the latest information from the official LIANA+ documentation

Key Dependencies

LIANA+ typically works with these complementary packages:

• scanpy: Single-cell analysis and data manipulation
• omnipath: Access to comprehensive prior knowledge databases
• decoupler: Pathway enrichment and transcription factor activity analysis
• plotnine: Grammar of graphics plotting in Python
• mudata: Multi-modal data structure support