METAFLUX

High-Performance Genome-Scale Metabolic Flux Analysis for Bulk and Single-Cell Transcriptomics

📖 Documentation: https://zaoqu-liu.github.io/METAFlux/

Abstract

METAFLUX is a computationally optimized R package for genome-scale metabolic flux analysis (MFA) from bulk and single-cell RNA sequencing data. Building upon the theoretical framework of METAFlux (Huang et al., Nature Communications, 2023), this implementation leverages the Human Genome-scale Metabolic model (Human-GEM) to infer 13,082 intracellular metabolic reaction fluxes through constraint-based flux balance analysis (FBA).

The package provides:

Metabolic Reaction Activity Scores (MRAS) derived from gene expression profiles
Flux Balance Analysis (FBA) via quadratic programming optimization
Community metabolic modeling for single-cell tumor microenvironment analysis
Cross-platform compatibility with optimized parallel computing

Theoretical Background

Flux Balance Analysis

METAFLUX employs constraint-based metabolic modeling to predict intracellular fluxes. The steady-state assumption yields:

$$\mathbf{S} \cdot \mathbf{v} = 0$$

where $\mathbf{S}$ is the stoichiometric matrix (8,378 metabolites × 13,082 reactions) and $\mathbf{v}$ is the flux vector. The optimization objective maximizes biomass production:

$$\max_{\mathbf{v}} \quad c^T \mathbf{v} \quad \text{subject to} \quad \mathbf{S} \cdot \mathbf{v} = 0, \quad \mathbf{lb} \leq \mathbf{v} \leq \mathbf{ub}$$

Gene-Protein-Reaction Mapping

Reaction activity scores integrate gene expression with enzyme kinetics through:

Isoenzyme (OR): $\text{Score} = \sum_i \frac{e_i}{w_i}$ (additive expression)
Enzyme complex (AND): $\text{Score} = \min_i \frac{e_i}{w_i}$ (rate-limiting subunit)

where $e_i$ is gene expression and $w_i$ is the gene occurrence weight across the metabolic network.

Community Modeling

For single-cell analysis, METAFLUX constructs a community stoichiometric matrix incorporating:

Intracellular metabolism within each cell type
Intercellular metabolite exchange
Competition for shared extracellular nutrients

Installation

From R-universe (Recommended)

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

From GitHub

# install.packages("remotes")
remotes::install_github("Zaoqu-Liu/METAFLUX")

Dependencies

# Core dependencies
install.packages(c("osqp", "Matrix", "dplyr", "stringr", "stringi"))

# For single-cell analysis
install.packages("Seurat")  # Version ≥ 4.0.0

# For parallel computing
install.packages(c("parallel", "foreach", "doParallel"))

Quick Start

Bulk RNA-seq Pipeline

library(METAFLUX)

# Load example data
data("bulk_test_example")  # Gene expression matrix
data("human_blood")        # Nutrient medium composition

# Step 1: Calculate Metabolic Reaction Activity Scores (MRAS)
mras <- calculate_reaction_score(bulk_test_example)

# Step 2: Compute metabolic fluxes via FBA
flux <- compute_flux(mras = mras, medium = human_blood)

# Analyze specific pathways
glycolysis_flux <- flux[grep("glycolysis", rownames(flux), ignore.case = TRUE), ]

Single-Cell RNA-seq Pipeline

library(METAFLUX)

# Load Seurat object and medium
data("sc_test_example")
data("human_blood")

# Step 1: Bootstrap aggregation for noise reduction
avg_expr <- calculate_avg_exp(
  myseurat = sc_test_example,
  myident = "Cell_type",
  n_bootstrap = 100,
  seed = 42
)

# Step 2: Calculate MRAS
mras <- calculate_reaction_score(avg_expr)

# Step 3: Define cell type fractions
fractions <- as.numeric(table(sc_test_example$Cell_type) / ncol(sc_test_example))

# Step 4: Community-based flux calculation
flux <- compute_sc_flux(
  num_cell = length(unique(sc_test_example$Cell_type)),
  fraction = fractions,
  fluxscore = mras,
  medium = human_blood,
  parallel = TRUE,
  n_cores = 4
)

Data Requirements

Input Data

Data Type	Format	Requirements
Bulk RNA-seq	Matrix/Data frame	Rows: HGNC gene symbols; Columns: samples; Values: normalized expression (TPM/FPKM)
Single-cell	Seurat object	RNA assay with normalized counts; metadata with cell type annotations
Medium	Data frame	Columns: `metabolite`, `reaction_name`

Available Medium Profiles

human_blood: Human serum/plasma composition (for tissue samples)
cell_medium: DMEM/F12 culture medium (for cell lines)

Output Interpretation

Flux Values

Value	Interpretation
Positive	Net production / forward reaction
Negative	Net consumption / uptake
Zero	Reaction inactive or at steady state

Key Reactions

# Biomass production (growth rate)
flux["biomass_human", ]

# Glucose uptake
flux["HMR_9034", ]  # Should be negative (consumption)

# Lactate secretion (Warburg effect)
flux["HMR_9135", ]  # Positive in cancer cells

Performance

Computational benchmarks on Intel i7-10700K (8 cores), 32GB RAM:

Analysis Type	Dataset Size	Execution Time
Bulk RNA-seq	100 samples	~2 min
Bulk RNA-seq	500 samples	~10 min
Single-cell	5,000 cells × 4 types	~4 min
Single-cell	10,000 cells × 8 types	~15 min

Citation

If you use METAFLUX in your research, please cite:

Huang Y, Mohanty V, Dede M, Tsai K, Daher M, Li L, Rezvani K, Chen K. Characterizing cancer metabolism from bulk and single-cell RNA-seq data using METAFlux. Nature Communications 14, 4883 (2023). https://doi.org/10.1038/s41467-023-40457-w

@article{huang2023metaflux,
  title={Characterizing cancer metabolism from bulk and single-cell RNA-seq data using METAFlux},
  author={Huang, Yuefan and Mohanty, Vakul and Dede, Merve and Tsai, Kyle and Daher, May and Li, Li and Rezvani, Katayoun and Chen, Ken},
  journal={Nature Communications},
  volume={14},
  number={1},
  pages={4883},
  year={2023},
  publisher={Nature Publishing Group},
  doi={10.1038/s41467-023-40457-w}
}

Related Resources

Original METAFlux: https://github.com/KChen-lab/METAFlux
Human-GEM: https://github.com/SysBioChalmers/Human-GEM
Tutorial: https://htmlpreview.github.io/?https://github.com/KChen-lab/METAFlux/blob/main/Tutorials/pipeline.html

License

Acknowledgments

This package builds upon the foundational work of the Ken Chen Laboratory at MD Anderson Cancer Center. We acknowledge the Human-GEM consortium for providing the genome-scale metabolic model.