Gene-level alignment of single-cell trajectories.

Area of Science:

• Computational Biology
• Genomics
• Systems Biology

Background:

• Single-cell data analysis enables the inference of dynamic cellular processes, such as development and response to stimuli, through pseudotime trajectory construction.
• Existing methods for trajectory comparison often rely on dynamic programming but struggle with assumptions of definitive matches, limiting their accuracy.

Purpose of the Study:

• To introduce Genes2Genes, a novel Bayesian information-theoretic dynamic programming framework designed for accurate alignment of single-cell trajectories.
• To enable the capture of sequential gene-level matches and mismatches between reference and query trajectories.

Main Methods:

• Developed a Bayesian information-theoretic dynamic programming framework named Genes2Genes.
• Implemented a method to analyze sequential matches and mismatches of individual genes between trajectories.
• Applied the framework to both simulated and real-world single-cell datasets, including disease cell-state trajectory analysis.

Main Results:

• Genes2Genes accurately inferred trajectory alignments across diverse datasets.
• The framework identified distinct clusters of alignment patterns, revealing nuanced relationships between trajectories.
• Demonstrated utility in analyzing disease cell-state trajectories and pinpointing divergences from in vivo systems.

Conclusions:

• Genes2Genes provides a robust and accurate method for aligning single-cell trajectories, overcoming limitations of previous approaches.
• The framework's ability to precisely align trajectories aids in understanding cellular dynamics and optimizing in vitro culture conditions by identifying deviations from in vivo states.