Metabolite Communication - Usage Guide

Overview

Metabolite-mediated cell-cell communication infers which cell types exchange metabolites by scoring enzyme-to-sensor expression from scRNA-seq (MEBOCOST), with flux (scFEA), FBA-state (Compass), and neurotransmitter (NeuronChat) alternatives. Metabolite levels are never measured - they are inferred through a chain of enzyme expression -> flux -> level -> sensing - so this is the most speculative communication layer and is hypothesis-generation only, requiring metabolomics, mass-spectrometry imaging, or tracing to validate.

Prerequisites

pip install mebocost scanpy anndata
# If the PyPI build is unavailable or stale, install from source:
# pip install git+https://github.com/kaifuchenlab/MEBOCOST
# MEBOCOST also needs its metabolite-enzyme-sensor database and a mebocost.conf
# pointing at those files (download from the MEBOCOST repository).

Quick Start

Tell your AI agent what you want to do: - "Score metabolite communication between my cell types with MEBOCOST" - "Which cell types express the machinery to secrete lactate, and which sense it?" - "Compare metabolic signaling between tumor and normal tissue" - "Separate the high-confidence receptor calls from the bidirectional-transporter calls"

Example Prompts

Basic Analysis

"Run MEBOCOST on my annotated scRNA-seq data and filter on permutation FDR"

"Find significant metabolite sender-receiver pairs between cell types"

Specific Metabolites

"Which cells express the enzymes for prostaglandin E2 and which express its sensors?"

"Show the inferred amino-acid signaling and flag the transporter-based calls"

Comparative Analysis

"Compare metabolite communication between treatment and control"

"Find metabolic signaling differences in the tumor microenvironment"

Validation Framing

"List orthogonal experiments to confirm the top metabolite communications"

What the Agent Will Do

1. Verify the data is log-normalized and uses gene symbols, and that ambient RNA was handled in preprocessing
2. Build a MEBOCOST object with cell-type labels and a config file pointing at the metabolite-sensor database
3. Run permutation inference and filter on the FDR, not the raw p-value
4. Summarize by metabolite and sender->receiver pair, separating lower-confidence transporter-based calls
5. State the enzyme->flux->level->sensing chain explicitly and frame every result as a hypothesis
6. Recommend orthogonal validation (metabolomics, MSI, isotope tracing, or enzyme/sensor perturbation)

Tips

• Double inference - Metabolite levels are inferred from enzyme expression, never measured; report "machinery consistent with producing X", never "cell A produces X".
• Filter on FDR - Use permutation_test_fdr, not the raw permutation p-value; significance is statistical, not a measured concentration.
• Capitalized columns - The result table uses Sender, Receiver, Metabolite_Name, Sensor, Annotation, Commu_Score, Norm_Commu_Score, permutation_test_fdr.
• Transporters are bidirectional - A transporter "sensor" may export rather than import and may move several metabolites, so sender/receiver direction can be wrong; treat transporter calls as lower confidence.
• Decontaminate first - Ambient RNA inflates enzyme and sensor expression; run SoupX/DecontX/CellBender before inference.
• Gene symbols and normalization - MEBOCOST needs gene symbols (not Ensembl IDs) and log-normalized data, plus at least ~10 cells per group for stable statistics.
• No spatial geometry - Metabolites diffuse and degrade, but dissociated data has no coordinates; do not claim neighbor exchange without spatial metabolomics.
• Pick the right tool - MEBOCOST for enzyme-sensor crosstalk, scFEA for per-cell flux, Compass for metabolic-state comparison, NeuronChat for neural systems; verify the database version and required files against the installed docs.
• Validation is non-optional - Confirm with targeted metabolomics, MALDI/DESI imaging, isotope tracing, or enzyme/sensor knockout; expression-only metabolite CCC alone is a lead, not a finding.

Related Skills

single-cell/cell-communication - Ligand-receptor CCC; the single-inference counterpart this skill mirrors at one extra remove single-cell/cell-annotation - Cell-type labels define metabolite senders and receivers single-cell/preprocessing - Log-normalization, gene-symbol mapping, and ambient-RNA decontamination happen here, before inference metabolomics/pathway-mapping - Places inferred metabolites in pathway context and informs which to prioritize metabolomics/isotope-tracing - Orthogonal flux validation that a producing cell actually makes the metabolite systems-biology/flux-balance-analysis - Genome-scale FBA underlying Compass-style per-cell metabolic state