Cell-Cell Communication

Overview

Cell-cell communication is fundamental to coordinated cardiac function. HeartMAP analyzes intercellular signaling by inferring ligand-receptor (L-R) interactions from single-cell RNA-seq co-expression patterns, revealing how different cell types communicate to maintain heart health and respond to disease.

Key Insight: Communication hubs in the heart include atrial cardiomyocytes and adipocytes with hub scores of 0.037 to 0.047, indicating their central role in coordinating cardiac signaling.

Ligand-Receptor Analysis

What Are Ligand-Receptor Interactions?

Ligand Production

A sender cell produces and secretes a signaling molecule (ligand):

Proteins (growth factors, cytokines)
Peptides (natriuretic peptides)
Small molecules

Receptor Binding

A receiver cell expresses a receptor protein that binds the ligand:

Transmembrane receptors
G-protein coupled receptors
Receptor tyrosine kinases

Signal Transduction

Binding triggers intracellular signaling cascades:

Gene expression changes
Metabolic shifts
Cellular responses

The L-R Database

HeartMAP uses curated ligand-receptor pair databases:

LIANA (Preferred)
Fallback Database

Ligand-receptor Inference FrameworkWhen available, HeartMAP uses LIANA’s consensus database:

‘consensus’: Curated from multiple sources (recommended)
‘cellphonedb’: CellPhoneDB database
‘omnipath’: OmniPath database
‘connectome’: Ramilowski 2015
‘cellinker’: CellLinker database

# LIANA integration
from heartmap import Config
config = Config.default()
config.analysis.use_liana = True

Key Cardiac L-R Interactions

Angiogenesis & Vascular

VEGF Family (Vascular Endothelial Growth Factor):

VEGFA → FLT1/KDR/NRP1: Blood vessel formation
VEGFC → FLT4: Lymphatic development
ANGPT1/2 → TEK: Vessel stabilization

Clinical Relevance: Coronary angiogenesis, therapeutic angiogenesis post-MI

TGF-beta Superfamily

Fibrosis and ECM Regulation:

TGFB1/2/3 → TGFBR1/TGFBR2: Fibroblast activation
BMP2/4/7 → BMPR1A/BMPR2: Cardiac development

Clinical Relevance: Cardiac fibrosis, remodeling after injury, heart failure progression

Inflammatory Signaling

Cytokines:

IL6 → IL6R: Inflammation, cardiac remodeling
TNF → TNFRSF1A/B: Inflammatory response, apoptosis
IL1B → IL1R1: Acute inflammation

Chemokines:

CXCL12 → CXCR4/CXCR7: Immune cell recruitment
CCL2 → CCR2: Monocyte infiltration

Clinical Relevance: Myocarditis, post-MI inflammation, chronic heart failure

Cardiac-Specific

Neuregulin Signaling:

NRG1 → ERBB2/ERBB4: Cardiomyocyte survival, contractility

Endothelin System:

EDN1 → EDNRA/EDNRB: Vasoconstriction, remodeling

Natriuretic Peptides:

NPPA/NPPB → NPR1: Blood pressure regulation, diuresis

Clinical Relevance: Heart failure therapies (BNP), cardiotoxicity prevention (NRG1)

Metabolic Signaling

Growth Factors:

IGF1 → IGF1R: Cardiomyocyte growth, metabolism
FGF2 → FGFR1/2: Angiogenesis, cardioprotection

Metabolic Hormones:

LEP → LEPR: Leptin signaling, metabolic regulation
ADIPOQ → ADIPOR1: Adiponectin, cardioprotection

Clinical Relevance: Metabolic syndrome, diabetic cardiomyopathy

Communication Scoring Method

HeartMAP calculates communication strength using co-expression analysis:

Algorithm

Calculate Mean Expression per Cell Type

For each cell type, compute mean expression of all genes:

# Pseudocode
for cell_type in celltypes:
    cells = adata[adata.obs['leiden'] == cell_type]
    mean_expr[cell_type] = cells.X.mean(axis=0)

Filter L-R Pairs

Keep only pairs where both genes are expressed:

# Filter by expression threshold (e.g., > 0.1)
if ligand_expr > 0.1 and receptor_expr > 0.1:
    # Proceed with scoring

Calculate Communication Score

Use geometric mean of co-expression:

# Communication strength formula
score = sqrt(ligand_expr × receptor_expr)

Rationale:

Requires both ligand AND receptor to be expressed
Balanced weighting of sender and receiver
Dampens false positives from one-sided expression

Build Communication Matrix

Create cell-type × cell-type interaction matrix:

DataFrame columns:
- source: sender cell type
- target: receiver cell type  
- ligand: ligand gene
- receptor: receptor gene
- communication_score: strength (0-1)

Example Calculation

# Real example from HeartMAP source code
# src/heartmap/pipelines/__init__.py:216-268

for ligand, receptor in ligand_receptor_pairs:
    if ligand not in adata.var_names or receptor not in adata.var_names:
        continue
    
    ligand_idx = list(adata.var_names).index(ligand)
    receptor_idx = list(adata.var_names).index(receptor)
    
    for source_type in cell_types:
        ligand_expr = cell_type_expression[source_type][ligand_idx]
        
        if ligand_expr > 0.1:  # Expression threshold
            for target_type in cell_types:
                if source_type == target_type:
                    continue  # Skip self-communication
                
                receptor_expr = cell_type_expression[target_type][receptor_idx]
                
                if receptor_expr > 0.1:
                    # Calculate strength
                    strength = np.sqrt(ligand_expr * receptor_expr)
                    
                    communication_data.append({
                        'source': source_type,
                        'target': target_type,
                        'ligand': ligand,
                        'receptor': receptor,
                        'communication_score': strength
                    })

Communication Hub Analysis

What Are Communication Hubs?

Cells that send or receive many signals, acting as coordination centers in the cardiac communication network.

High Hub Score

Characteristics:

Express many ligands/receptors
Connect to many other cell types
Central to network topology
Potential therapeutic targets

Low Hub Score

Characteristics:

Specialized, limited communication
Peripheral network position
Focused signaling roles

Hub Score Calculation

HeartMAP uses expression variability to identify hubs:

# Hub score formula (from source code)
hub_score = (std × mean) / (variance + 1)

Components:

std: Standard deviation of expression (variability)
mean: Mean expression (overall level)
variance: Expression variance (normalized denominator)

Interpretation:

High variability + high expression = high hub score
Indicates dynamic signaling capacity
Normalized to prevent outlier dominance

Known Cardiac Hubs

Atrial Cardiomyocytes
Adipocytes
Endothelial Cells
Fibroblasts

Hub Score: 0.037 - 0.047Key Signaling:

NPPA/NPPB secretion (natriuretic peptides)
EDN1 production (endothelin)
Mechanical stretch sensing
Endocrine function

Why High Hub Score:

Central to hemodynamic regulation
Communicate with endothelial, fibroblasts, immune cells
Respond to volume/pressure changes

Pathway Analysis

HeartMAP identifies enriched signaling pathways in cardiac communication:

Cardiac Development Pathways

Notch signaling: DLL1/4, JAG1/2 → NOTCH1/2/3
Wnt signaling: WNT3A, WNT5A → FZD receptors
BMP/TGF-beta: Chamber formation, valve development

Hemodynamic Regulation

Natriuretic peptide pathway: NPPA/NPPB → NPR1
Endothelin system: EDN1 → EDNRA/EDNRB
Adrenergic signaling: Indirect (receptor expression)

Angiogenesis & ECM

VEGF pathway: VEGFA/C → FLT1/KDR/FLT4
FGF pathway: FGF2/7/9/10 → FGFR1/2/3
Integrin-ECM: Collagen, fibronectin, laminin interactions

Inflammation & Immune

Cytokine signaling: IL6, TNF, IL1B pathways
Chemokine gradients: CXCL12, CCL2, CCL5
Complement system: C3, C5 signaling

Running Communication Analysis

Basic Usage

from heartmap import Config
from heartmap.pipelines import AdvancedCommunicationPipeline

# Initialize configuration
config = Config.default()
config.analysis.use_liana = True  # Use LIANA if available

# Create pipeline
pipeline = AdvancedCommunicationPipeline(config)

# Run analysis (requires annotated data)
results = pipeline.run(
    data_path='results/basic/annotated_data.h5ad',
    output_dir='results/communication'
)

# Access results
comm_scores = results['results']['communication_scores']
hub_scores = results['results']['hub_scores']
pathway_scores = results['results']['pathway_scores']

print(f"Detected {len(comm_scores)} L-R interactions")
print(f"\nTop 5 interactions:")
print(comm_scores.nlargest(5, 'communication_score'))

Interpreting Results

Communication Scores
Hub Scores
Pathway Scores

DataFrame format:

# communication_scores.csv
source          target          ligand  receptor  communication_score
Cardiomyocyte  Fibroblast      TGFB1   TGFBR1    0.856
Endothelial    Cardiomyocyte   VEGFA   KDR       0.742
Fibroblast     Endothelial     FN1     ITGA5     0.689
...

Key columns:

source: Cell type sending signal
target: Cell type receiving signal
ligand: Ligand gene name
receptor: Receptor gene name
communication_score: Strength (0-1)

Series format (per cell):

# hub_scores.csv
cell_barcode         hub_score
AAACCCAAGAAACACT-1   0.042
AAACCCAAGAAACCAT-1   0.038
AAACCCAAGAAACCGC-1   0.015
...

Interpretation:

Scores range 0-1 (typically 0.01-0.05 for hubs)
Higher = more central to communication network
Can aggregate by cell type for cell-type hub scores

Visualizations

Automatically generated plots:

Communication Heatmap

Cell-type × cell-type matrix showing interaction strength

Hub Score Plot

UMAP colored by hub scores, highlighting communication centers

Pathway Scores

Bar plot of enriched pathways (when available)

Advanced: Custom L-R Analysis

Using Custom L-R Database

from heartmap.data import LigandReceptorDatabase
import scanpy as sc

# Load custom L-R pairs
db = LigandReceptorDatabase(resource='consensus')  # or 'cellphonedb'

# Filter to genes in your data
adata = sc.read_h5ad('annotated_data.h5ad')
available_genes = adata.var_names.tolist()

lr_pairs = db.get_pairs(
    confidence_threshold=0.8,  # High confidence only
    present_in_data=available_genes
)

print(f"Found {len(lr_pairs)} high-confidence L-R pairs in dataset")

# Export for review
db.save_to_csv('my_lr_database.csv')

Filtering by Specific Pathways

import pandas as pd

# Load communication results
comm = pd.read_csv('results/communication/communication_scores.csv')

# Filter for angiogenesis pathway
vegf_interactions = comm[
    comm['ligand'].str.contains('VEGF') | 
    comm['receptor'].str.contains('FLT|KDR')
]

print("VEGF pathway interactions:")
print(vegf_interactions)

# Filter for inflammatory signaling
inflam_ligands = ['IL6', 'TNF', 'IL1B', 'IFNG']
inflam_comm = comm[comm['ligand'].isin(inflam_ligands)]

Chamber-Specific Communication

Combine communication and chamber analysis:

# Requires chamber labels in data
adata = sc.read_h5ad('multi_chamber_annotated.h5ad')

# Subset to specific chamber
lv_cells = adata[adata.obs['chamber'] == 'LV']

# Run communication analysis on LV only
from heartmap.pipelines import AdvancedCommunicationPipeline
pipeline = AdvancedCommunicationPipeline(config)
lv_results = pipeline.run(lv_cells, 'results/lv_communication')

print("LV-specific communication hubs:")
print(lv_results['results']['hub_scores'].nlargest(10))

Clinical Applications

Drug Target Identification

Use communication hubs and pathways:

Block pathological signaling (TGF-beta in fibrosis)
Enhance protective signaling (NRG1-ERBB2)
Target cell-type specific receptors

Disease Mechanism Understanding

Compare healthy vs. diseased:

Altered communication networks in heart failure
Inflammatory pathway activation in myocarditis
Dysregulated angiogenesis post-MI

Biomarker Discovery

Communication-derived biomarkers:

NPPA/NPPB (already clinical: BNP test)
Ligands detectable in blood
Predict response to therapy

Precision Medicine

Personalized signaling maps:

Patient-specific communication networks
Targeted therapy selection
Predict drug response

Next Steps

Chamber Analysis

Combine communication with chamber-specific analysis

API Reference

Detailed API documentation

References

Method Reference:

LIANA: Dimitrov et al. (2022). Nature Methods.
CellPhoneDB: Efremova et al. (2020). Nature Protocols.

HeartMAP Application: Kgabeng, T., et al. (2025). HeartMAP: A Multi-Chamber Spatial Framework for Cardiac Cell-Cell Communication. Computational and Structural Biotechnology Journal.

Get Started

Core Concepts

Guides

Examples

​Overview

​Ligand-Receptor Analysis

​What Are Ligand-Receptor Interactions?

​The L-R Database

​Key Cardiac L-R Interactions

​Communication Scoring Method

​Algorithm

​Example Calculation

​Communication Hub Analysis

​What Are Communication Hubs?

High Hub Score

Low Hub Score

​Hub Score Calculation

​Known Cardiac Hubs

​Pathway Analysis

​Running Communication Analysis

​Basic Usage

​Interpreting Results

​Visualizations

Communication Heatmap

Hub Score Plot

Pathway Scores

​Advanced: Custom L-R Analysis

​Using Custom L-R Database

​Filtering by Specific Pathways

​Chamber-Specific Communication

​Clinical Applications

Drug Target Identification

Disease Mechanism Understanding

Biomarker Discovery

Precision Medicine

​Next Steps

Chamber Analysis

API Reference

​References

Build docs developers (and LLMs) love

Overview

Ligand-Receptor Analysis

What Are Ligand-Receptor Interactions?

The L-R Database

Key Cardiac L-R Interactions

Communication Scoring Method

Algorithm

Example Calculation

Communication Hub Analysis

What Are Communication Hubs?

Hub Score Calculation

Known Cardiac Hubs

Pathway Analysis

Running Communication Analysis

Basic Usage

Interpreting Results

Visualizations

Advanced: Custom L-R Analysis

Using Custom L-R Database

Filtering by Specific Pathways

Chamber-Specific Communication

Clinical Applications

Next Steps

References