Welcome to ReCoN’s documentation!¶

ReCoN is a new tool for reconstructing multicellular models.

It combines both gene regulatory networks and cell communication networks to explore the molecular coordinations between multiple cell types — all at once.

ReCoN uses heterogeneous multilayer networks and integrates several layers of information into a complex network, ready to be explored and analyzed. Both the GRNs and intercellular networks are inferred from single-cell RNA-seq data (and optionally scATAC-seq).

Note

💡 The philosophy behind ReCoN:
🧬 Cells do not act in isolation, but in a coordinated, dynamic system.

ReCoN can be used to address several biological questions, including:

ReCoN use cases

📦 Installation¶

ReCoN is available as a Python package and can be installed through pip.

conda create -n recon python=3.10
conda activate recon
pip install recon[grn-lite]

If you are generating your grn externally, you can install ReCoN without the GRN dependencies.:raw-html:<br /> You should then be able to use more recent version of Python.

pip install recon

⚠️ To generate GRNs, ReCoN requires CellOracle to be installed. Since CellOracle requires quite old dependency versions, we propose to install our own lite branch that contains only the necessary functions through the code “recon[grn-lite]”.

Installation troubleshooting & Frequently asked questions

See the Troubleshooting and FAQs page

💊 How does a treatment affect the molecular state of multiple cell types?¶

ReCoN can be used to predict how a treatment (e.g., a drug) affects the molecular state of each cell type in a multicellular context (e.g., organ, tumor microenvironment).
It represents both direct effect, through treatment - receptor bindings, and indirect effects of the treatment, through cell communication interplays.

ReCoN-indirect-effect Remi-Trimbour 2025

Note

A treatment effect can be decomposed into two components:

Direct effect — The effect of the treatment caused by direct binding of the receptors of a cell type.
Indirect effect — The effect on a cell type mediated by other cell types that respond and, in turn, secrete ligands which modulate the molecular state of the cell type of interest.

ReCoN models these two components through different random walk with restart (RWR) processes on the multicellular network (cf. ReCoN overview and algorithm). The parameter \(\alpha \in [0, 1]\) allows to tune the relative importance of the direct and indirect effects. \(\alpha\) is the weight of the direct effect, while \(1 - \alpha\) is the weight of the indirect effect.

ReCoN-direct-indirect-effect-formula Remi-Trimbour 2025

Why indirect effects matter

Importance of indirect effects

Surrounding cells can secrete ligands in response to the treatment, which then feed back and alter signaling and regulation in the focal cell type.

In our evaluation of ReCoN, we found that giving more importance to the indirect effect (\(\alpha = 0.8\) — or an indirect effect 4 times stronger than the direct effect) led to the best performance in both individual and multiple perturbation setups.
(Trimbour et al., 2025 — Immune Dictionary and Heart Failure showcases)

See how to use ReCoN to model the effect of a drug here : Predicting treatment effects in multicellular systems.

🧫 Understanding multicellular program coordination¶

How are the surrounding cells regulating and impacted by the state of a given cell-type ? ReCoN can be used to explore the causes and consequences of a given cell/multicellular state.

We can then identify the key molecules and cell types that are involved in the coordination of this state.

ReCoN-multicellular-programs Remi-Trimbour 2025

See how to use ReCoN to explore multicellular coordination here : Exploring multicellular coordination.

⚙️ Visualizing multicellular molecular cascades modulating cell states¶

ReCoN can reconstruct the intercellular cascades driving a specific transcriptomic state. These cascades includes intracellular elements (receptors, transcription factors), but also ligands and their own regulators. It offers you a comprehensive view of these interactions, and the possibility to identify new potential targets at different regulatory levels.

See how to use ReCoN to explore intracellular regulatory elements here : Exploring molecular cascades and identify regulators.

🧬 Building GRNs through HuMMuS methodology¶

We previously developped HuMMuS (Trimbour et al., 2024), an other method based on heterogeneous multilayer networks to build gene regulatory networks from single-cell RNA-seq and ATAC-seq data. HuMMuS can be used as a standalone method, but it was initially developed as an hybrid package with R and Python.

ReCoN can now also be used to run a Python implementation of HuMMuS to infer GRNs from single-cell data. This implementation leverages the functions of CellOracle to build the prior knowledge links between TF, DNA regions and target genes. HuMMuS is then applied on a multilayer composed of a TF layer, a DNA region layer and a target gene layer to infer the final GRN.

See how to use ReCoN to build GRNs with HuMMuS here : Building gene regulatory networks with HuMMuS.

📖 Cite ReCoN¶

If you use ReCoN in your work, please cite:

Cite ReCoN

Trimbour R., Ramirez Flores R. O., Saez Rodriguez J., Cantini L. (2026). Modelling multicellular coordination by bridging cell-cell communication and intracellular regulation through multilayer networks. bioRxiv. https://doi.org/10.64898/2026.01.20.700561

If you also use ReCoN to generate GRNs, please cite:

Cite HuMMuS

Trimbour R., Deutschmann I. M., Cantini L. (2024). HuMMuS: Inferring gene regulatory networks through heterogeneous multilayer networks. Bioinformatics, 40(3), btae143. https://doi.org/10.1093/bioinformatics/btae143

Note

To download the tutorial data, use the following commands:

All tutorial data:

from recon.data import fetch_all_tutorial_data
fetch_all_tutorial_data(data_dir='./data')

Specific file (e.g., RNA data):

from recon.data import fetch_tutorial_data
fetch_tutorial_data('perturbation_tuto/rna.h5ad', data_dir='./data')