DensityPath: an algorithm to visualize and reconstruct cell state-transition path on density landscape for...

Area of Science:

• Computational biology
• Single-cell genomics
• Bioinformatics

Background:

• Analyzing single-cell RNA sequencing (scRNA-seq) data presents challenges in visualizing high-dimensional expression profiles.
• Reconstructing cell developmental trajectories from static scRNA-seq snapshots is computationally complex.

Purpose of the Study:

• To develop an algorithm, DensityPath, for visualizing scRNA-seq data structure and reconstructing cell state-transition paths.
• To effectively handle high dimensionality and heterogeneity inherent in scRNA-seq datasets.

Main Methods:

• Utilizing elastic embedding for non-linear dimension reduction to preserve data structure.
• Employing a single-cell multimodal density landscape to identify representative cell states via topological features.
• Reconstructing cell trajectories by finding the geodesic minimum spanning tree on the density landscape.

Main Results:

• DensityPath successfully visualizes intrinsic data structures in a 2D embedded space.
• The algorithm reconstructs complex cell developmental trajectories, including those with multiple bifurcations.
• DensityPath demonstrates high accuracy in pseudotime calculation and branch assignment for both real and simulated scRNA-seq data.
• The method is robust to parameter variations and data permutations.

Conclusions:

• DensityPath offers an efficient and accurate approach for analyzing cell developmental trajectories from scRNA-seq data.
• The algorithm provides a powerful tool for understanding cell fate decisions and transcriptional heterogeneity.
• DensityPath software is publicly available for broader research applications.