Setting the SCENE for Interpretable Cell-Gene Embeddings in Single-Cell RNA-seq
Abstract
Single-cell RNA sequencing measures cellular states at high resolution, but sparse high-dimensional count data remain difficult to model interpretably. We introduce the Single-Cell Euclidean Network Embedding (SCENE), a probabilistic latent-distance model that jointly embeds cells and genes from Unique Molecular Identifier (UMI) counts. SCENE treats the count matrix as a weighted bipartite cell-gene graph, where Euclidean distances represent transcriptional affinity, and combines this geometry with a zero-inflated count likelihood that separates gene detection from expression magnitude. Across real and simulated scRNA-seq datasets, SCENE recovers biologically structured cell and gene embeddings with state-of-the-art performance. Surprisingly, major biological structure is preserved in native two- and three-dimensional latent spaces, enabling directly interpretable visualization. Perturbation analyses show that SCENE organizes glucocorticoid-response genes and T-cell receptor regulatory programs coherently in gene space, capturing biology beyond cell-type separation. SCENE provides a transparent representation learning framework in which low-dimensional Euclidean geometry supports accurate modeling and biological interpretation.
Related
- DECONVersation: Single Cell Foundation Model-Derived Embeddings for Robust Cell Type Deconvolution of Bulk RNA-seq
- Deep generative embeddings of gene expression and splicing reposition the interpretation of single-cell transcriptomic signatures
- SpCAST enables scalable and interpretable integration of single-cell RNA sequencing and single-cell-resolved spatial transcriptomics
- Ground-truth evaluation of sparse autoencoder features on spatial, single-cell and language-model embeddings
- Extracting interpretable single-cell metabolic states with graph-guided representation learning
BibTeX
@misc{moberg2026setting,
title = {{Setting the SCENE for Interpretable Cell-Gene Embeddings in Single-Cell RNA-seq}},
author = {O. L. Moberg and M. B. Petersen and T. Herlau and L. E. Kristensen and L. E. Jessen and M. Morup},
howpublished = {bioRxiv},
year = {2026},
month = sep,
doi = {10.64898/2026.09.12.750699},
url = {https://www.biorxiv.org/content/10.64898/2026.09.12.750699v1},
note = {CC BY},
}
RIS
TY - UNPB
TI - Setting the SCENE for Interpretable Cell-Gene Embeddings in Single-Cell RNA-seq
AU - O. L. Moberg
AU - M. B. Petersen
AU - T. Herlau
AU - L. E. Kristensen
AU - L. E. Jessen
AU - M. Morup
PY - 2026
DA - 2026/09/12
PB - bioRxiv
DO - 10.64898/2026.09.12.750699
UR - https://www.biorxiv.org/content/10.64898/2026.09.12.750699v1
AB - Single-cell RNA sequencing measures cellular states at high resolution, but sparse high-dimensional count data remain difficult to model interpretably. We introduce the Single-Cell Euclidean Network Embedding (SCENE), a probabilistic latent-distance model that jointly embeds cells and genes from Unique Molecular Identifier (UMI) counts. SCENE treats the count matrix as a weighted bipartite cell-gene graph, where Euclidean distances represent transcriptional affinity, and combines this geometry with a zero-inflated count likelihood that separates gene detection from expression magnitude. Across real and simulated scRNA-seq datasets, SCENE recovers biologically structured cell and gene embeddings with state-of-the-art performance. Surprisingly, major biological structure is preserved in native two- and three-dimensional latent spaces, enabling directly interpretable visualization. Perturbation analyses show that SCENE organizes glucocorticoid-response genes and T-cell receptor regulatory programs coherently in gene space, capturing biology beyond cell-type separation. SCENE provides a transparent representation learning framework in which low-dimensional Euclidean geometry supports accurate modeling and biological interpretation.
ER -