CellUntangler: Separating distinct biological signals in single-cell data with deep generative models

TL;DR

CellUntangler is a deep learning tool that separates overlapping biological signals in single-cell RNA data, such as cell cycle and cell type.

Contribution

CellUntangler introduces a deep generative model with multiple latent subspaces to disentangle complex biological signals in single-cell data.

Findings

CellUntangler successfully disentangles the cell cycle from other processes in both cycling-only and mixed datasets.

The model generalizes to separate additional signals like spatial zonation and interferon response.

It identifies marker genes associated with specific biological signals of interest.

Abstract

Single-cell RNA sequencing has provided new insights into both intracellular and intercellular processes. However, multiple processes, such as cell-type programs, differentiation, and the cell cycle, often occur simultaneously within one cell. Existing methods typically target a single process and impose restrictive assumptions, risking the loss of valuable biological information. We introduce CellUntangler, a deep generative model that embeds cells into a latent space composed of multiple subspaces, each tailored with an appropriate geometry to capture a distinct signal. Applied to datasets of cycling-only and mixed cycling/non-cycling cells, CellUntangler disentangles the cell cycle from other processes such as cell type. The framework generalizes to disentangle additional signals, including spatial, tissue dissociation, interferon response, and cell-type identity. By providing flexible embeddings to capture various signals, CellUntangler enables selective enhancement or filtering of signals at the gene-expression level, offering a powerful tool for disentangling complex biological processes in single-cell data.

•CellUntangler disentangles multiple biological signals in scRNA-seq data•Applicable to diverse signals such as the cell cycle and spatial zonation•Captures cell-cycle dynamics in datasets with cycling-only and mixed cells•Identifies marker genes associated with signals of interest

CellUntangler disentangles multiple biological signals in scRNA-seq data

Applicable to diverse signals such as the cell cycle and spatial zonation

Captures cell-cycle dynamics in datasets with cycling-only and mixed cells

Identifies marker genes associated with signals of interest

Chen et al. present CellUntangler, a deep-learning model that utilizes a flexible latent space composed of multiple subspaces, allowing for the capture and filtering of biological signals in single-cell RNA sequencing data. CellUntangler demonstrates broad extensibility across a wide range of signals.

Figures