Latent periodic process inference from single-cell RNA-seq data.

Area of Science:

• Genomics
• Computational Biology
• Developmental Biology

Background:

• Multicellular organism development involves complex biological processes.
• Single-cell RNA sequencing (scRNA-seq) enables inference of developmental trajectories.
• Accurate characterization of periodic processes like the cell cycle remains challenging.

Purpose of the Study:

• To develop a robust computational method for identifying circular trajectories in gene expression data.
• To improve the accuracy and robustness of cell cycle phase determination from scRNA-seq data.
• To enhance the utility of scRNA-seq in biological studies by mitigating cell cycle effects.

Main Methods:

• Development of Cyclum, an autoencoder-based computational approach.
• Identification of circular trajectories in gene expression space using Cyclum.
• Application of Cyclum for cell cycle effect removal in scRNA-seq datasets.

Main Results:

• Cyclum demonstrates substantial improvements in accuracy and robustness for cell cycle characterization compared to existing methods.
• Removal of cell cycle effects using Cyclum significantly enhances the delineation of cell subpopulations.
• Improved cell subpopulation identification facilitates the construction of cell atlases and analysis of tumor heterogeneity.

Conclusions:

• Cyclum is an effective tool for characterizing periodic biological processes, specifically the cell cycle, in scRNA-seq data.
• Mitigating cell cycle artifacts with Cyclum improves downstream analyses, including cell subpopulation discovery.
• The method has broad applications in cell atlas projects and the study of cancer biology.