GPseudoClust: deconvolution of shared pseudo-profiles at single-cell resolution.

Area of Science:

• Computational biology
• Genomics
• Bioinformatics

Background:

• Traditional gene clustering methods struggle with single-cell RNA sequencing data due to unobserved temporal ordering.
• Existing pseudotime methods for ordering cells introduce uncertainty, potentially leading to inaccurate gene clusters.

Purpose of the Study:

• To develop a novel method, GPseudoClust, that simultaneously infers gene clusters and cell pseudotemporal order.
• To quantify the uncertainty associated with both pseudotemporal ordering and gene clustering.

Main Methods:

• GPseudoClust integrates a recent pseudotime inference method with non-parametric Bayesian clustering.
• Efficient Markov Chain Monte Carlo sampling and subsampling strategies are employed for computational efficiency.
• The method is validated on diverse simulated and experimental single-cell RNA sequencing datasets.

Main Results:

• GPseudoClust effectively infers gene clusters and pseudotemporal ordering simultaneously.
• The method accurately quantifies uncertainty in both inferred ordering and clustering.
• Demonstrated effectiveness across various simulated and real-world single-cell datasets.

Conclusions:

• GPseudoClust provides a robust framework for analyzing time-course single-cell gene expression data.
• Accounting for uncertainty in pseudotime is crucial for reliable gene clustering.
• The developed method enhances the accuracy and interpretability of single-cell trajectory inference.