destiny

destiny implements diffusion map spectral methods to perform nonlinear dimensionality reduction and visualize single-cell expression data, preserving manifold geometry to analyze cellular heterogeneity and dynamic processes.

Key Features:

• R implementation: Provides an efficient implementation within the R programming environment leveraging spectral methods for diffusion maps.
• Diffusion map algorithm: Employs the diffusion map algorithm to capture the intrinsic geometry of high-dimensional single-cell expression data.
• Single-cell-specific noise model: Incorporates a noise model that handles missing and censored values common in single-cell datasets.
• Efficient nearest-neighbour approximation: Uses an efficient nearest-neighbour approximation to scale analyses to datasets comprising hundreds of thousands of cells.
• Projection functionality: Includes functions to project new data onto precomputed diffusion maps for integration and comparison.

Scientific Applications:

• Single-cell heterogeneity and dynamics: Characterizes cellular heterogeneity and dynamic processes from single-cell expression profiles.
• Time-resolved mass cytometry and reprogramming: Applied to time-resolved mass cytometry datasets to track developmental trajectories and cellular reprogramming.

Methodology:

Implements the diffusion map algorithm (spectral methods) with a single-cell-specific noise model, an efficient nearest-neighbour approximation, and projection functions for mapping new data onto existing diffusion maps.