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

Area of Science:

• Computational Biology
• Genomics
• Single-cell analysis

Background:

• Single-cell RNA sequencing (scRNA-seq) reveals cellular complexity but struggles with simultaneous biological processes.
• Existing methods often oversimplify by focusing on single processes, potentially losing critical biological information.
• Disentangling concurrent signals like cell type and cell cycle is crucial for accurate biological interpretation.

Purpose of the Study:

• To develop a novel deep generative model, CellUntangler, for disentangling multiple biological signals within single-cell data.
• To address the limitations of existing methods in handling simultaneous cellular processes.
• To provide a flexible framework for analyzing complex single-cell gene expression data.

Main Methods:

• Introduced CellUntangler, a deep generative model utilizing a latent space with multiple subspaces.
• Each subspace is geometrically tailored to capture distinct biological signals.
• Applied the model to scRNA-seq datasets, including those with and without cell cycle activity.

Main Results:

• CellUntangler successfully disentangled the cell cycle from other processes, such as cell type.
• The framework demonstrated generalizability in disentangling additional signals like spatial information, tissue dissociation effects, interferon response, and cell-type identity.
• The model enables selective enhancement or filtering of signals at the gene-expression level.

Conclusions:

• CellUntangler offers a powerful and flexible tool for dissecting complex biological processes in single-cell data.
• The ability to disentangle multiple signals improves the accuracy and depth of scRNA-seq data analysis.
• This approach facilitates a more nuanced understanding of cellular heterogeneity and function.