scMetaLink

scMetaLink is an R package for inferring metabolite-mediated cell-cell communication from single-cell and spatial transcriptomics data.

📖 Documentation: https://Zaoqu-Liu.github.io/scMetaLink/

Background

Metabolites serve as critical signaling molecules in the cellular microenvironment. While tools like CellChat and CellPhoneDB have advanced our understanding of ligand-receptor interactions, the metabolic dimension of intercellular communication remains underexplored. scMetaLink addresses this gap by integrating enzyme expression, transporter activity, and receptor-metabolite interactions to reconstruct metabolite-mediated communication networks.

Workflow

Features

• Metabolite Production Inference: Estimate metabolite production potential based on biosynthetic enzyme expression
• Metabolite Sensing Analysis: Quantify sensing capability via receptors (GPCRs, nuclear receptors) and transporters
• Statistical Framework: Permutation-based significance testing with multiple hypothesis correction
• Pathway Integration: Aggregate communication patterns at the pathway level
• Spatial Transcriptomics: Distance-weighted communication analysis for spatial data
• Visualization: Heatmaps, chord diagrams, network plots, and spatial maps

Database

scMetaLink utilizes MetalinksDB, containing:

• 41,894 metabolite-protein interactions
• 1,128 metabolites
• 4,374 proteins/genes
• 157,741 pathway associations

Installation

From r-universe (Recommended)

install.packages("scMetaLink", repos = "https://zaoqu-liu.r-universe.dev")

From GitHub

# Install devtools if needed
if (!require("devtools")) install.packages("devtools")

devtools::install_github("Zaoqu-Liu/scMetaLink")

Quick Start

Single-Cell Analysis

library(scMetaLink)

# Create scMetaLink object
obj <- createScMetaLink(
  expression_data = expr_matrix,
  cell_meta = cell_metadata,
  cell_type_column = "cell_type"
)

# Infer metabolite production and sensing
obj <- inferProduction(obj)
obj <- inferSensing(obj)

# Compute communication with significance testing
obj <- computeCommunication(obj, n_permutations = 1000)
obj <- filterSignificantInteractions(obj, pvalue_threshold = 0.05)

# Visualize results
plotCommunicationHeatmap(obj)
plotCommunicationCircle(obj)

For Seurat users:

result <- runScMetaLinkSeurat(seurat_obj, cell_type_column = "cell_type")

---

Spatial Transcriptomics Analysis

scMetaLink supports spatial transcriptomics data with distance-weighted communication analysis.

Spatial Workflow

Spatial Communication Algorithm

Spatial Quick Start

library(scMetaLink)

# Load spatial data
data(st_expr)
data(st_meta)
data(st_scalefactors)

# Create spatial scMetaLink object
obj <- createScMetaLinkFromSpatial(
  expression_data = st_expr,
  spatial_coords = st_meta[, c("x", "y")],
  cell_meta = st_meta,
  cell_type_column = "cell_type",
  scale_factors = st_scalefactors
)

# Infer production and sensing
obj <- inferProduction(obj)
obj <- inferSensing(obj)

# Compute spatially-weighted communication
obj <- computeSpatialCommunication(
  obj,
  method = "knn",        # K-nearest neighbors (recommended for Visium)
  k_neighbors = 6,       # Hexagonal grid neighbors
  n_permutations = 1000
)

# Filter and visualize
obj <- filterSignificantInteractions(obj)
plotSpatialCellTypes(obj)
plotSpatialCommunicationNetwork(obj)
plotSpatialFeature(obj, metabolite = "L-Lactic acid", type = "production")

Spatial Weighting Methods

Method	Formula	Best for
knn	K-nearest neighbors	Visium (recommended)
gaussian	exp(-d²/2σ²)	Smooth decay
exponential	exp(-d/λ)	Sharp decay
threshold	Binary cutoff	Strict distance limit

---

Methodology

Metabolite Production Potential (MPP)

$$MPP(m, c) = \frac{\sum_{e \in E_m^+} w_e \cdot \text{expr}(e, c)}{|E_m^+|} \cdot S(m)$$

where $E_m^+$ denotes biosynthetic enzymes for metabolite $m$, $w_e$ is the evidence weight, and $S(m)$ represents secretion potential.

Metabolite Sensing Capability (MSC)

$$MSC(m, c) = \sum_{r \in R_m} w_r \cdot \tau(r) \cdot \text{expr}(r, c)$$

where $R_m$ includes receptors and transporters for metabolite $m$, and $\tau(r)$ is the receptor type weight.

Communication Score

$$C(s \rightarrow r, m) = \sqrt{MPP(m, s) \times MSC(m, r)}$$

Spatial Communication Score

$$C_{spatial}(s \rightarrow r, m) = \sqrt{MPP(m, s) \times MSC(m, r)} \times w(d_{sr})$$

where $w(d)$ is the spatial weight function (e.g., Gaussian decay).

Statistical significance is assessed via permutation testing with FDR correction.

---

Tutorials

Tutorial	Description
Quick Start	Basic workflow
Theory	Mathematical framework
Production & Sensing	Inference details
Communication	Statistical analysis
Spatial Analysis	Spatial transcriptomics
Lactate Signaling	Lactate-mediated communication
Visualization	Plotting functions
Applications	Case studies

Citation

If you use scMetaLink in your research, please cite:

Liu Z, et al. (2026). scMetaLink: Inferring metabolite-mediated cell-cell
communication from single-cell and spatial transcriptomics data.

Contact

• Author: Zaoqu Liu
• Email: liuzaoqu@163.com
• GitHub: https://github.com/Zaoqu-Liu/scMetaLink