GPseudoRank: a permutation sampler for single cell orderings.

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Pseudotime methods for single-cell RNA sequencing (scRNA-seq) data typically provide point estimates for cell ordering.
• Modeling the uncertainty of pseudotime estimates remains a challenge.
• There is a need for methods that can sample from complex, multi-modal cell ordering distributions and quantify uncertainty changes during biological development.

Purpose of the Study:

• To introduce GPseudoRank, a novel method for modeling pseudotime uncertainty in scRNA-seq data.
• To enable sampling from complex and multi-modal posterior distributions of cell orders.
• To estimate dynamic changes in ordering uncertainty throughout biological processes.

Main Methods:

• GPseudoRank utilizes Bayesian modeling and Markov Chain Monte Carlo (MCMC) methods.
• The approach is designed to handle complex and multi-modal posterior distributions of cell orders.
• A variant of the method is developed for scalability to large droplet-based scRNA-seq datasets.

Main Results:

• GPseudoRank successfully samples from complex posterior distributions and identifies phases of varying pseudotime uncertainty.
• The method accurately identifies cells with early antiviral responses.
• Pseudotime uncertainty is linked to the identification of metastable states in biological processes.

Conclusions:

• GPseudoRank offers a robust framework for analyzing pseudotime uncertainty in scRNA-seq data.
• The method provides insights into dynamic changes in developmental trajectories and cellular states.
• GPseudoRank facilitates downstream analyses such as gene clustering and network inference by quantifying ordering uncertainty.