DTFLOW

DTFLOW infers pseudo-temporal cellular trajectories from single-cell RNA sequencing data to characterize multi-branching differentiation processes.

Key Features:

• Bhattacharyya Kernel Feature Decomposition (BKFD): Performs dimension reduction by establishing a stationary distribution per cell that represents transitions between cellular developmental states.
• Random walk with restart algorithm: Models transitions between cellular states within the BKFD framework.
• Bhattacharyya kernel matrix distance metric: Computes distances used to calculate pseudo-times of single cells.
• Reverse Searching on kNN Graph (RSKG): Traces back through nearest-neighbor relationships on a kNN graph to identify multi-branching differentiation processes.
• Multi-branching pseudo-temporal inference: Integrates BKFD-derived distances and RSKG-based graph traversal to infer branching pseudo-time trajectories.

Scientific Applications:

• Cellular Differentiation Studies: Infers temporal ordering and branching of cells transitioning from progenitor to differentiated states using single-cell RNA sequencing data.
• Developmental Biology Research: Reveals multi-branching developmental pathways at single-cell resolution.
• Disease Progression Analysis: Maps cellular transition trajectories to analyze progression pathways in disease contexts.

Methodology:

Performs dimension reduction via Bhattacharyya Kernel Feature Decomposition (BKFD) that establishes a stationary distribution per cell and uses a random walk with restart to model transitions; computes pseudo-times using a Bhattacharyya kernel matrix-based distance metric; employs Reverse Searching on kNN Graph (RSKG) to trace nearest-neighbor paths and identify multi-branching processes.