Pseudotime graph diffusion for post hoc visualization of inferred single-cell trajectories.

Area of Science:

• Computational biology
• Single-cell analysis
• Bioinformatics

Background:

• Visualizations are crucial for interpreting single-cell trajectories but often fail to accurately represent trajectory structure.
• Inaccurate trajectory representations compromise the interpretation of cellular dynamics and downstream analyses.

Purpose of the Study:

• To introduce Pseudotime Graph Diffusion (PGD), a novel post hoc framework for enhancing the interpretation of single-cell trajectory data.
• To improve the fidelity of visual representations of inferred cellular trajectories.

Main Methods:

• PGD employs random-walk diffusion on a pseudotime graph to smooth cell-level features.
• The framework propagates information along inferred trajectory paths to enhance continuity and structure.
• PGD was applied to visualize monocyte and macrophage trajectories during wound healing.

Main Results:

• PGD-smoothed embeddings significantly improved the visualization of complex inferred trajectories.
• The method demonstrated effectiveness in smoothing monocyte and macrophage dynamics during wound healing.
• PGD was extended to enable trajectory-aware gene expression smoothing.

Conclusions:

• PGD offers a lightweight and interpretable solution for improving single-cell trajectory analysis.
• By enhancing the agreement between visualizations and inferred trajectories, PGD facilitates more accurate interpretation of dynamic cellular processes.
• The framework provides a valuable tool for researchers studying cellular dynamics and differentiation pathways.